Working Paper No.

75

An endogenous growth model with public capital

and sustainable government debt

by

Alfred Greiner

University of Bielefeld
Department of Economics
Center for Empirical Macroeconomics
P.O. Box 100 131
33501 Bielefeld, Germany
An endogenous growth model with public capital and
sustainable government debt

Alfred Greiner∗

Abstract

This paper presents and analyzes an endogenous growth model with public cap-
ital and public debt. It is assumed that the ratio of the primary surplus to gross
domestic income is a positive linear function of the debt income ratio which assures
that public debt is sustainable. The paper then derives necessary conditions for the
existence of a sustainable balanced growth path for the analytical model. Further,
simulations are undertaken in order to gain insights into stability properties of the
model and in order to analyze growth effects of deficit financed increases in public
investment. The latter is done for the model on the sustainable balanced growth
path as well as for the model along the transition path.

JEL: E62, H60, H54

Keywords: Balanced Growth, Transition Dynamics, Limit Cycles, Fiscal Policy

∗
Department of Economics, Bielefeld University, P.O. Box 100131, 33501 Bielefeld, Germany
1 Introduction
One strand in endogenous growth theory postulates that sustained per-capita growth
results from ongoing investment in public capital which raises the incentive of private
investors to build up a private capital stock. Productive public capital has a long tradition
in the economics literature. Arrow and Kurz [1] were among the first to present a formal
model with that type of capital. However, their approach did not allow for sustained
per-capita growth in the long-run. Futagami et al. [10] then presented an endogenous
growth model with productive public capital which generates sustained per-capita growth
in the long-run. Their model basically is a more general version of the simple approach
presented by Barro [4]. The difference between these two models is that Futagami et al.
assume that public investment does not affect aggregate production possibilities directly,
as does Barro, but only indirectly by building up a stock of public capital which stimulates
economic production.
One consequence of the model presented by Futagami et al. is that their model gives
rise to transition dynamics, which does not hold for the Barro model. Both models, how-
ever, have in common that the budget of the government is balanced at any moment in
time, as frequently assumed in this class of models. Further, as a consequence of pro-
ductive public capital there exists a growth maximizing income tax rate with an inverted
U-relationship between the balanced growth rate and the size of the income tax rate.
In Greiner and Hanusch [13] the model by Futagami et al. is extended and allows for
both productive and non-productive public spending and it is demonstrated that growth
and welfare maximization may be different even if one confines the investigation to the
balanced growth path.
In the approaches mentioned above the public capital stock is a purely public good
which is non-rivalrous and non-excludable. Glomm and Ravikumar [12] in their review
of the literature present a model where this issue is addressed among others. Further,
they explicitly distinguish between government expenditures which enter as inputs in the

1
production function for output and expenditures which raise the productivity of invest-
ment technologies. Baier and Glomm [3] extend the approach by allowing for an elasticity
of substitution between public and private capital which is not necessarily equal to one.
These authors demonstrate that the elasticity of substitution affects the growth maximiz-
ing ratio between private and public capital and, as a consequence, the growth maximizing
tax rates on capital and labour.
As concerns the empirical relevance of public capital for the productivity of economies
the results are not unambiguous. A frequently cited study is the paper by Aschauer [2], for
example, who reports strong effects of public capital. Further, he states that public capital
is dramatically more important than public investment as a flow variable. However, there
are also studies which reach different conclusions. This is not too surprising because it is
to be expected that the time period under consideration as well as the countries which are
considered are important as to the results obtained. For a survey of the empirical studies
dealing with public spending, public capital and the economic performance of countries
see Sturmet al. [20] and Pfähler et al. [19].
All what the theoretical models have in common is that they assume a balanced
government budget. An exception to this assumption is provided by the model presented
by Turnovsky [22], chap. 13, who allowed for public debt in his analysis. He demonstrates
that an increase in public investment financed by higher public debt unambiguously raises
the balanced growth rate (p. 418). The reason for that outcome is that public capital
stimulates investment and public debt does not affect the allocation of resources in the
long-run and, therefore, does not have negative growth effects.
On the other hand, public debt and the question of whether public debt is sustainable
plays an important role in market economies. So, the latter question has been the subject
of a great many empirical studies, in particular as concerns the U.S. (see e.g. Hamilton
and Flavin [15], Kremers [17], Wilcox [24], or Trehan and Walsh [21]). However, no un-
ambiguous answer could be obtained and Bohn [7], [8] criticized these tests because they

2
make assumptions about future states of nature that are difficult to estimate from a single
set of observed time series data. Therefore, he proposes a different test which analyzes
whether the ratio of the primary surplus to gross domestic product is a positive linear
function of the ratio of public debt to gross domestic product which guarantees sustain-
ability of public debt. The reasoning behind this argument is that if a government raises
the primary surplus as public debt increases it takes corrective actions which stabilize the
debt ratio. This implies that the debt ratio displays mean-reversion and thus the ratio
remains bounded implying that public debt is sustainable.
The empirical analysis for the U.S. indeed confirms that a higher debt ratio leads to
higher primary surpluses (cf. Bohn [8]). The same also holds for countries in the EURO
area (see Greiner et al. [14]). Thus, the intertemporal budget constraint, although it
should be fulfilled only in infinity, has immediate repercussions for the period budget
constraint since the government reduces public spending or/and raises tax revenues as
public debt rises.
In this paper, we present a theoretical model where we combine the two topics men-
tioned above. That is we present an endogenous growth with public investment following
the approach by Futagami et al. [10] and we integrate public debt. Further, we assume
that the primary surplus of the government is a positive linear function of public debt
which guarantees that the intertemporal budget constraint of the government holds. Given
this assumption the paper then analyzes the structure of model where we pay particular
attention to the dynamic behaviour.
The rest of the paper is organized as follows. In the next section we demonstrate that
sustainability of public debt is given if the primary surplus is a positive linear function
of public debt. In section 3 we present the endogenous growth model with public capital
and government debt. Section 4 studies the implications of the model and analyzes
growth effects of deficit financed increases in public investment both for the model on the
sustainable growth path and taking into account transition dynamics. Section 5, finally,

3
concludes the paper.

2 The primary surplus and sustainability of public

debt
The accounting identity describing the accumulation of public debt in continuous time is
given by:
Ḃ(t) = B(t)r(t) − S(t), (1)

where B(t) stands for real public debt,1 r(t) is the real interest rate, and S(t) is real
government surplus exclusive of interest payments.
Solving equation (1) we get for the level of public debt at time t
Rt
 Z t R 
τ
r(τ )dτ r(µ)dµ
B(t) = e 0 B(0) − e 0
−
S(τ )dτ , (2)
0
Rt
r(τ )dτ
with B(0) public debt at time t. Multiplying both sides of (2) with e− 0 , to get the
present value of government debt at time t, yields
R
Z t R
− 0t r(τ )dτ τ
e B(t) + e− 0 r(µ)dµ S(τ )dτ = B(0). (3)
0

If the first term in (3) goes to zero in the limit the current value of public debt equals
the sum of discounted future non-interest surpluses. Then, we have
Z t R
τ
B(0) = e− 0 r(µ)dµ S(τ )dτ. (4)
0

Equation (4) is the present-value borrowing constraint and we call a path of public debt
which satisfies this constraint a sustainable debt. It states that public debt at time zero
must equal the future present-value surpluses. Equivalent to requiring that (4) must be
fulfilled is that the following condition holds:
Rt
r(τ )dτ
lim e− 0 B(t) = 0. (5)
t→∞

1
Strictly speaking, B(t) should be real public net debt.

4
That equation is usually referred to as the no-Ponzi game condition (see e.g. Blanchard
and Fischer (1989), ch. 2).
Now, assume that the ratio of the primary surplus to gross domestic income ratio is
a positive linear function of the debt to gross domestic income ratio and of a constant.
The primary surplus ratio, then, can be written as

T (t) − Ip (t) B(t)

=φ+β , (6)
Y (t) Y (t)

where T (t) denotes the tax revenue at time t, Ip (t) is public spending at t, Y (t) gross
domestic income at t and φ, β ∈ IR are constants. All variables are real variables. It
should be noted that β determines how strong the primary surplus reacts to changes in
public debt and, therefore, can be considered as a feedback parameter of public debt. φ
determines whether the level of the primary surplus rises or falls with an increase in gross
domestic income.
Using that equation the differential equation describing the evolution of public debt
can be written as

Ḃ(t) = r(t) B(t) − T (t) + Ip (t) = (r(t) − β) B(t) − φ Y (t). (7)

Rt
r(τ )dτ
Solving this differential equation and multiplying both sides with e− 0 to get the
present value of public debt yields
R R
 Z t 
− 0t r(τ )dτ βτ − 0τ (r(µ)−γy (µ))dµ
e B(t) = e −βt
B(0) − φ Y (0) e dτ , (8)
0

with B(0) public debt at time t = 0 and γy the growth rate of gross domestic income.
First, we state that for r < γy the intertemporal budget constraint is irrelevant because
in this case the economy is dynamically inefficient implying that the government can play
a Ponzi game. Therefore, we only consider the case r > γy .
Writing equation (8) as
Rt Rτ
Rt (r(µ)−γy (µ))dµ
r(τ )dτ 0
eβτ e− 0 dτ
e− 0 B(t) = e −βt
B(0) − φ Y (0) . (9)
eβ t
5
 Rt
r(τ )dτ
shows that β > 0 is a necessary condition for limt→∞ e− 0 B(t), i.e. for the present
value of public debt to converge to zero for t → ∞.
Rτ
If the numerator in the second expression in (9) remains finite, implying that 0
(r(µ)−
γy (µ))dµ converges to infinity, the second term converges to zero. If the numerator in the
Rt
second expression in (9) becomes infinite, l’Hôpital gives the limit as e− 0 (r(µ)−γy (µ))dµ /β.
Rt
This shows that β > 0 and limt→∞ 0 (r(µ)−γy (µ))dµ = ∞ are sufficient for sustainability
of public debt.
These considerations demonstrate that the intertemporal budget constraint of the
government is fulfilled if the ratio of the primary surplus to gross domestic income is a
positive linear function of the debt ratio, which can also be observed for economies in the
real world. Therefore, we posit that the government sets the primary surplus according
to (6) implying that public debt is sustainable. In the next section, we present our
endogenous growth model with public capital and with that assumption.

3 The structure of the growth model

Our economy consists of three sectors: A household sector which receives labour income
and income from its saving, a productive sector and the government. First, we describe
the household and the productive sector.

3.1 The household and the productive sector

The household sector is represented by one household which maximizes the discounted
stream of utility resulting from per-capita consumption, C,2 over an infinite time horizon
subject to its budget constraint. The utility function is assumed to be logarithmic, U (C) =
ln C, and the household has one unit of labour, L, which it supplies inelastically. The
2
From now on we omit the time argument t if no ambiguity arises.

6
maximization problem, then, can be written as
Z ∞
max e−ρt ln C dt, (10)
C 0

subject to
(1 − τ ) (w + rW + π) = Ẇ + C. (11)

ρ is the subjective discount rate, w is the wage rate, r is the interest rate and π gives
possible profits from the productive sector which the household takes as given in solving
its optimization problem. W ≡ B + K denotes assets which are equal to public debt, B,
and private capital, K. All variables give per-capita quantities. τ ∈ (0, 1) is the income
tax rate. The dot gives the derivative with respect to time and we neglect depreciation
of private capital.
To solve this problem we formulate the present-value Hamiltonian which is written as

H = ln C + λ((1 − τ ) (w + rW + π) − C) (12)

Necessary optimality conditions are given by

C −1 = λ (13)

λ̇ = ρλ − λ(1 − τ ) r (14)

If the transversality condition limt→∞ e−ρt W/C = 0 holds which is fulfilled for a time path
on which assets grow at the same rate as consumption the necessary conditions are also
sufficient.
The productive sector is represented by one firm which behaves competitively and
which maximizes static profits. The production function of the firm is given

Q = K 1−α Gα Lξ , (15)

with α + ξ ≤ 1. (1 − α) is the private capital share, α gives the public capital share and
ξ is the labour share. G denotes public capital which is assumed to be a purely public

7
good. Using that labour is normalized to one profit maximization yields

w = ξK 1−α Gα (16)

r = (1 − α)K −α Gα (17)

Resorting to (13), (14) and (16), (17), which must hold in equilibrium, the growth rate
of consumption is derived as

Ċ
= −ρ + (1 − τ )(1 − α)K −α Gα . (18)
C

3.1.1 The government

The government in our economy receives tax revenues from income taxation and has
revenues from issuing government bonds it then uses for public investment. Further, the
primary surplus is a positive linear function of public debt which guarantees that public
debt is sustainable as shown in the previous section.
Using (6) and defining ip ≡ (1 − φ/τ ), the budget constraint of the government can
be written as
Ḃ + T = rB + Ip ↔ Ḃ = (r − β)B + T (ip − 1) , (19)

with β > 0 and Ip = ip T − βB public investment which amounts to total public spending.
It should be noted that we have assumed that ip denotes that fraction of the tax revenue
the government uses for gross public investment. ip < 1 implies that a certain part of the
tax revenue is used for the debt service and ip > 1 implies that no part of the tax revenue
is used for the debt service and public investment may exceed the tax revenue.
Neglecting depreciation, public capital evolves according to

Ġ = Ip = ip T − βB . (20)

8
3.2 Equilibrium conditions and the sustainable balanced growth
path

An equilibrium allocation is defined as an allocation such that the firm maximizes profits
implying that factor prices equal their marginal products (equations (16) and (17)), the
household solves (10) subject to (11) and the budget constraint of the government (19) is
fulfilled.
The economy-wide resource constraint is obtained by combining equations (11) and
(19) as
K 1−α Gα
 
K̇ C T B
=− + − ip −β . (21)
K K K K K
Thus, the economy is completely described by equations (18), (19), (20) and (21) plus the
limiting transversality condition of the household.
A sustainable balanced growth path (SBGP) is defined as a path on which all endoge-
nous variables grow at the same rate, i.e. K̇/K = Ġ/G = Ḃ/B = Ċ/C holds, and the
intertemporal budget constraint of the government is fulfilled, that is equation (5) must
hold. Note that the SBGP is dynamically efficient3 and the transversality condition of
the household is fulfilled. Since we have posited that the government sets the primary
surplus according to (6) with β > 0 any path which satisfies K̇/K = Ġ/G = Ḃ/B = Ċ/C
is associated with a sustainable public debt. To make this clear we speak of a sustainable
balanced growth path.
To analyze our economy around a SBGP we define the new variables x ≡ G/K,
b ≡ B/K and c ≡ C/K. Differentiating these variables with respect to time yields a
3
The difference between the interest rate and the growth on the SBGP is strictly positive and constant
Rt
implying limt→∞ 0 (r(µ) − γy (µ))dµ = ∞.

9
three dimensional system of differential equations given by

ẋ = x c − βb(1 + x−1 ) − xα + (1 + (1 − α)b)ip τ xα (1 + x−1 ) ,

(22)

ḃ = b c − β(1 + b) + (1 − α)xα + (ip − 1)τ xα ((1 − α) + b−1 ) − xα +

ip τ xα (1 + (1 − α)b)) , (23)

ċ = c (c − ρ + (1 − τ )(1 − α)xα − xα − βb + ip τ xα (1 + (1 − α)b)) . (24)

A solution of ẋ = ḃ = ċ = 0 with respect to x, b, c gives a SBGP for our model and

the corresponding ratios x⋆ , b⋆ , c⋆ on the SBGP.4 In the next section we first analyze
the structure of our model and, then, investigate how deficit financed increases in public
investment affect the balanced growth rate and the growth rate on the transition path.

4 Implications of the model

To get insight into our model we first solve (24) with respect to c and insert that value
in (23) giving

ḃ
= (ρ − β) + (ip − 1)τ xα ((1 − α) + b−1 ) + (1 − α)τ xα . (25)
b

From equation (25) we can derive a first result.

Assume that the government is a debtor, that is b > 0 holds. Then, for β ≤ ρ the right
hand side in that equation can become zero, which is necessary for a SBGP to exist, only
if ip < 1. ip < 1 is given if the level of the primary surplus rises with an increase in gross
domestic income, i.e. φ > 0 must holds From an economic point of view this implies that a
certain part of the tax revenue must be used for the debt service if the economy is to grow
over time at a constant rate if β is relatively small, that is if the primary surplus does
not increase sufficiently as public debt rises. Relatively small means that the parameter
β is lower than the rate of time preference ρ. But this result only holds if the government
4 ⋆
The denotes SBGP values and we exclude the economically meaningless SBGP x⋆ = b⋆ = c⋆ = 0.

10
is a debtor, i.e. for b > 0. This is obvious because there is no need for the government to
reduce a possibly existing primary deficit if the government is a net lender.
For β > ρ, a SBGP can exist for ip > 1 and a positive public debt. ip > 1 implies that
the level of the primary surplus negatively depends on gross domestic income, i.e. φ < 0
holds. In this case, the reaction of the government to a higher debt ratio, modelled by the
parameter β, is sufficiently strong so that an increase in may go along with a reduction
in the primary surplus.
These considerations have given some first insights into our model. In the next sub-
section we will further pursue the question of whether a SBGP exists and whether it is
stable. In addition, we will analyze growth effects of deficit financed increases in public
investment for the model on the SBGP. With deficit financed increase in public invest-
ment we mean an increase in public investment, modelled by a rise in ip , which does not
go along with a higher income tax rate.

4.1 The economy on the SBGP

To analyze our model further, we resort to simulations. We do so because the analytical

model turns out to become too complex to derive further results. As a benchmark for
our simulations we set the income tax rate to ten percent, i.e. τ = 0.1, the elasticity of
production with respect to public capital is set to 25 percent, i.e. α = 0.25.5 The rate of
time preference is set to 30 percent, ρ = 0.3. Interpreting one time period as 3 (5, 10)
years then gives an annual rate of time preference of 10 (6, 3) percent.
In table 1 we report results of our simulations for values of β which are smaller than
the rate of time preference ρ. γ denotes the balanced growth rate and unstable means
that at least two eigenvalues are positive or have positive real parts.

5
For a survey of empirical studies giving estimates for that parameter see Pfähler [19] or Sturm [20].

11
 Table 1

β = 0.15 β = 0.25

ip b⋆ x⋆ γ Stability b⋆ x⋆ γ Stability
ip = 1.15 -0.05 0.4 0.238 unstable -0.1 0.45 0.252 unstable
ip = 1.05 -0.02 0.37 0.226 unstable -0.04 0.39 0.232 unstable
ip = 0.9 0.04 0.31 0.204 unstable 0.07 0.27 0.186 unstable
ip = 0.75 0.09 0.24 0.174 unstable 0.16 0.05 0.022 unstable
0.15 0.04 0 unstable
ip = 0.45 0.15 0.04 0 unstable no SBGP for ip ≤ 0.74
0.11 0.01 -0.088 unstable
no SBGP for ip ≤ 0.44

Table 1 confirms the result derived for the analytical model that for β < ρ a certain
part of the tax revenue must be used for the debt service, i.e. ip < 1 must hold, to get
sustained growth if public debt is positive. For ip > 1 sustained growth goes along with
a negative government debt, that is the government must be a creditor.
Further, one realizes that the smaller ip , i.e. the smaller that part of the tax revenue
used for public investment, the smaller is the balanced growth rate γ in case where the
SBGP is unique. This implies that raising public investment increases the balanced growth
rate. If that part of the tax revenue which is used for public investment, ip , falls below a
certain critical value the model does not yield sustained growth at all. This critical value
is the larger the larger the parameter β. From an economic point of view this is obvious
because a high β implies that a given level of public debt goes along with a low level of
public investment since a large fraction of public revenues is used for the debt service.
As to stability, the SBGP is unstable in all cases. The eigenvalues are real with two
being positive and one being negative. This means that there exists a one dimensional
stable manifold. If one takes x(0) and b(0) as given this implies that the set of initial

12
conditions {x(0), b(0), c(0)} lying on the stable manifold has Lebesgue measure zero. In
this case the economy can converge to the SBGP in the long-run only if the government
levies a lump-sum tax at t = 0 which is used to control B(0) implying that B(0), and
thus b(0), can be set. B(0) and C(0), then, must be chosen such that b(0) and c(0) lie
on the stable manifold and these values are uniquely determined. In addition, for small
values of ip two SBGPs exist where one goes along with a zero or negative growth rate,
respectively.6
To gain further insight into our model we next set β > ρ. The results of the simulations
are shown in table 2.

Table 2

β = 0.35 β = 0.4

ip b⋆ x⋆ γ Stability b⋆ x⋆ γ Stability
ip = 1.15 -0.47 0.72 0.322 unstable 0.16 0.1 0.077 stable
ip = 1.05 -0.24 0.56 0.283 unstable 0.08 0.26 0.181 stable
0.15 0.04 0 unstable
ip = 1.02 -0.13 0.47 0.258 unstable 0.04 0.31 0.205 stable
0.16 0.1 0.08 stable
ip = 0.9 no SBGP for ip < 1 -0.2 0.53 0.275 stable
ip = 0 no SBGP for ip < 1 -1.01 1.06 0.384 stable

To interpret table 2 we first consider the case β = 0.35.

One can see that, as in table 1, the balanced growth rate is the larger the larger the
fraction of the tax revenue which is used for public investment, ip , for the unstable SBGP.
Further, there exists a critical value for ip below which sustained growth is not feasible.
This critical value is larger than in table 1 because of the higher β. In addition, for about
6
A negative growth rate of public capital does not make sense in our model. It would imply that
public capital is sold to the private sector.

13
ip ∈ (1, 1.07) there exist two SBGPs. The unstable SBGP has one negative and two
positive real eigenvalues, the stable SBGP has one positive real eigenvalue and a pair of
complex conjugate eigenvalues with negative real parts. The balanced growth rate of the
stable SBGP negatively depends on ip in contrast to the growth rate associated with the
unstable SBGP which depends positively on ip .
As to stability, the stable SBGP looses stability with a rising value of ip . For about
ip ∈ (1, 1.028) the SBGP is stable and for ip > 1.028 the stable SBGP becomes unstable
and this SBGP disappears for ip > 1.07 leaving only the unstable SBGP. For ip = icrit
p =
1.028651 the stable SBGP undergoes a Hopf bifurcation and leads to unstable limit cycles.7
For a slightly different value of ρ, namely for ρ = 0.32,8 a supercritical Hopf bifurcation
can be observed for ip = icrit
p = 0.971824 which leads to stable limit cycles. In this case,
there exists an interval of ip with strictly positive measure for which the economy does
not converge to the SBGP but converges to persistent cycles. The limit cycles occur for
values of ip larger icrit
p . From an economic point of view this means that the economy is

characterized by sustained fluctuations around the SBGP. Figure 1 shows the limit cycle
in the (x−b−c) phase space where the orientation is counter clockwise as indicated by
the arrows.
7
For those computations we used the software LOCBIF, see Khibnik et al. [18], and MATCONT, see
Dhooge et al. [9].
8
With ρ = 0.32 there exist two SBGPs for about ip ∈ (0.96, 1).

14
 0.4933

0.4932

0.4931
c

0.493

0.4929

0.1572

0.0894
0.0892
0.089
b 0.0888
0.157 0.0886
0.0884
0.0882
x

Figure 1: Limit cycle in the (x − b − c) phase space.

To understand the emergence of limit cycles from an economic point of view, we assume
that the economy originally is on the SBGP. The government, then, raises ip such that
this parameter falls in that interval of ip which generates stable cycles. As a consequence
of the increase in ip , public investment rises leading to an increase in the growth rate of
public capital and in the ratio G/K = x. The latter increase raises the marginal product
of private capital and leads to a higher growth rate of private consumption and of the
ratio C/K = c. As a result of the increase in ip , however, public debt also rises implying
that the resources for the debt service increase leading to a rise in the ratio B/K = b.
From figure 1 it can be seen that b lags behind x which makes sense from an economic
point of view. The increase in resources required for the debt service, finally, leads to
a decrease in the growth rate of public capital and in the ratio G/K = x. The latter
effect reduces the marginal product of private capital and leads to a smaller growth rate

15
of private consumption and to a decline in the ratio C/K = c. When the public debt ratio
has fallen enough public investment rises again which spurs economic growth.
In this way, a cyclical evolution is generated. It should be noted that for lower values of
β these fluctuations cannot be observed because then β is not sufficiently high to stabilize
the economy. For larger values of β cycles are excluded, too, because high values of β
tend to stabilize the economy in a way that it always converges to the SBGP as we will
see next.
Next, we consider the case β = 0.4.
Table 2 shows that, in this case, there exists a unique SBGP which is stable. The
high value of β guarantees that the primary surplus of the government reacts sufficiently
strong to higher public debt which stabilizes the economy. The eigenvalues are real with
two being negative and one being positive implying that there exists a two-dimensional
stable manifold and a unique c(0) so that the economy converges to the SBGP in the
long-run. Now, however, the balanced growth rate negatively depends on ip . In this case,
a deficit financed increase in public investment is offset by the higher public debt, which
requires more resources for the debt service, so that the economy finally invests less in
public capital. Sustained growth is also given for small values of ip and even for ip = 0.
But again, the government must be a creditor in this case. For ip = 0 public investment is
completely financed by public wealth which increases over time due to interest payments
and due to the tax revenue. The same outcome is observed for β = 0.5.
Before we go on with our analysis we briefly summarize our results obtained from the
simulations up to now. We saw that the higher β, i.e. the stronger the primary surplus
and, thus public spending, react to increases in public debt, the sooner the model is stable.
In this case, a deficit financed increase in public investment reduces the balanced growth
rate because of the strong feedback effects associated with public debt. It could also be
shown that the model may be very sensitive with respect to β, the feedback parameter
of public debt, and with respect to ip , giving that part of the tax revenue used for public

16
investment. So, variations in these two parameters may lead to stable limit cycles and
multiple SBGPs.
Further, for small values of β there exists a critical value of ip below which sustained
growth is not possible. This critical value is the large the larger is β which makes sense
from an economic point of view. Finally, it could be realized that for β > ρ that part of the
tax revenue used for public investment, ip , must be smaller than one to achieve sustained
growth unless the government is a creditor. A fact already shown for the analytical model.

Robustness of the results

Before we study the model along the transition path we investigate whether changing the
numerical parameter values affects the qualitative outcome. To do so we first set α = 0.15
and leave τ and ρ unchanged. Then, we set α = 0.15 and τ = 0.3 and leave ρ unchanged.
Finally, we change all three parameters and set α = 0.15, τ = 0.3 and ρ = 0.03.
Performing the calculations which led to tables 1 and 2 with the other parameter
values shows that the qualitative results do not change.9 That is, for small values of β
the model is unstable and there exists a critical value for ip below which sustained growth
is not feasible. The balanced growth rate associated with this SBGP positively depends
on ip , the fraction of the tax revenue used for public investment.
For higher values of β two SBGPs exist with one being unstable and the other being
stable. Further, the stable SBGP looses stability as ip is increased and becomes unstable
before it vanishes when ip is further increased. When β is increased further there exists
a unique SBGP which is stable. The balanced growth rate associated with this SBGP
negatively depends on ip .
Although the qualitative outcome does not change the values of β and ip which generate
the respective results and the numerical values of the endogenous variables are different.
Of course, this was to be expected. However, since we are interested in qualitative features
9
Tables reporting the exact outcome are available on request.

17
of our model this is of less importance.

4.2 Fiscal policy on the transition path

In this subsection we study the behaviour of our model after a once-and-for all deficit
financed increase in public investment taking into account transition dynamics. We as-
sume that the economy is originally on the SBGP when this fiscal policy is performed at
time t = 0. Further, we consider the case where our model is characterized by a saddle
point with two negative real eigenvalues.
To analyze the effects of a deficit financed increase in public investment we study the
solution of the linearized system of (22)-(24) which is given by

x(t) = x⋆ + C1 v11 eµ1 t + C2 v21 eµ2 t , (26)

b(t) = b⋆ + C1 v12 eµ1 t + C2 v22 eµ2 t , (27)

c(t) = c⋆ + C1 v13 eµ1 t + C2 v23 eµ2 t , (28)

with vij the j−th element of the eigenvector belonging to the negative real eigenvalue µi ,
i = 1, 2. Ci , i = 1, 2, are constants determined by the initial conditions x0 and b0 . Setting
t = 0 gives Ci , i = 1, 2, as a function of x0 and b0 . Inserting these Ci , i = 1, 2, in (28) gives
the unique c(0) on the stable manifold leading to the SBGP in the long-run. Given x(t),
b(t) and c(t) one can compute the growth rates of C, B, G and K according to (18)-(21).
To analyze growth effects of a deficit financed increase in public investment we take
the numerical example from the last subsection with β = 0.4 and raise ip from ip = 1.05
to ip = 1.055 which reduces the long-run balanced growth rate as can be seen from the
right part of table 2. The stable manifold of the linearized system and the adjustment to
the new SBGP are shown in Figure 2

18
 *
E

0.487

0.486
c 0.09
0.485
0.484 0.0875

0.25 0.085
b
0.2525

0.255 0.0825
x
0.2575
0.08
0.26

Figure 2: The stable manifold and the transition path to the new SBGP denoted by E.

At time t = 0, the ratio c jumps from the old SBGP value 0.4849 to c(0) = 0.4848
onto the stable manifold and then rises. Over time the path approaches the new SBGP
given by E = (x⋆ , b⋆ , c⋆ ) = (0.2497, 0.0882, 0.4855). One can see that x first rises and then
declines while b monotonically rises. c monotonically rises for t > 0, i.e. after the initial
downward adjustment at t = 0.
The reaction of the growth rates of private capital (solid line) and of public capital
(dotted line) to the increase in ip are shown in figure 3. For t < 0, the solid line gives the
balanced growth rate before the increase in ip .

19
 0.182

0.181

t
-20 20 40 60

0.179

0.178

Figure 3: Ġ/G (dotted line) and K̇/K (solid line) on the transition path.

First, we state that x and b do not react to parameter changes at time t = 0 since K, G
and B are predetermined variables which react only gradually while C immediately reacts
so that c jumps to the stable manifold. From figure 3 one realizes that an increase in ip
raises the growth rate of public capital, Ġ/G, which jumps upward at t = 0. (20) shows
that a rise in ip must raise Ġ/G at t = 0 since B as well as K and G and, thus the tax
revenue T , are fixed at t = 0. From an economic point of view this reaction is obvious since
raising the part of the tax revenue used for investment will cause an immediate increase
in public investment. Over time, however, the initial increase in public investment caused
by the higher ip is offset by the increase in public debt and, consequently, the growth rate
of public capital declines and approaches the new SBGP. So, figure 3 demonstrates that
both growth rates overshoot the long-run balanced growth rate, i.e. they first increase
before they decline.
As concerns the growth rate of private capital there are two counteracting effects. On
the one hand, the increase in ip reduces the consumption share c at t = 0, which tends

20
to raise the growth rate of private capital. On the other hand, the increase in the public
deficit, i.e. the rise in Ḃ, implies that there is a crowding-out of private saving at t = 0
which can be seen from (11). This effect tends to lower K̇/K. Figure 3 shows that the
crowding-out effect dominates implying that the growth rate of private capital declines
at t = 0. Over time, however, the growth rate of private capital then first rises before
it declines and approaches the new SBGP. The temporary rise in K̇/K results from the
increase in the public debt ratio, b, and from the temporary increase in production relative
to capital, xα , which imply a positive income effect for the private household. But for t
sufficiently large K̇/K falls because x declines over time.
It should be noticed that the rise in ip , which does not have distortions per se, affects
the growth rates of public and of private capital at t = 0 through the income effect. So,
the income effect, which is of no relevance as concerns the long-run balanced growth rate,
very well affects the growth rates of economic variables along the transition path.
Figure 4 shows the effects of the increase in ip on the growth rates of public debt
(dotted line) and of private consumption (solid line) where the solid line for t < 0 again
gives the balanced growth rate before the increase in ip .

21
 0.186

0.184

0.182

t
-20 20 40 60

0.178

Figure 4: Ḃ/B (dotted line) and Ċ/C (solid line) on the transition path.

Figure 4 shows that the growth rate of public debt increases at t = 0, due to the deficit
financed rise of public investment, and then declines and approaches its new SBGP value.
From (18) we see that the growth rate of consumption does not react to the increase
in ip at t = 0. Over time, Ċ/C first rises before it declines. The reason is that, as a
consequence of the increase in public investment, the ratio G/K = x first rises and then
declines implying the same effect for Ċ/C. Thus, as in the case of Ġ/G and K̇/K, the
growth rates of public debt and of private consumption first rise and then decline and
approach the SBGP value implying an overshooting of the long-run balanced growth rate.
These considerations show that the income effect associated with an increase in public
investment affects the growth rates of economic variables on the transition path. This
income effect generates an overshooting of the growth rates over the long-run balanced
growth rate.

22
5 Conclusion
This paper has presented an endogenous growth model with public capital and government
debt where the government raises the primary surplus as a result of higher public debt. In
dynamic efficient economies, the latter is sufficient for sustainability of public debt so that
any path on which all variables grow at the same rate can be called a sustainable balanced
growth path. The assumption that the primary surplus is a positive function of public
debt is also motivated by empirical studies (see the papers cited in the Introduction)
which present evidence that governments raise the ratio of the primary surplus to gross
domestic income as the debt ratio increases.
With this assumption the analysis of our endogenous growth model produced outcomes
which are different from those known in the literature. In particular, the following results
could be derived.
1. If the government is a debtor, it turned out that a sustainable balanced growth
path only exists if the government uses a certain part of the tax revenue for the debt
service, ip < 1, in case the primary surplus does not react sufficiently strong to higher
public debt, i.e. for β ≤ ρ. If the increase in the primary surplus is sufficiently strong as
public debt rises, β > ρ, sustained growth is feasible even if no part of the tax revenue is
used for the debt service, i.e. for ip > 1.
2. Numerical examples demonstrated that the sustainable balanced growth path is
very sensitive with respect to the fraction of the tax revenue used for public investment
and with respect to the parameter β determining the reaction of the primary surplus to a
rise in public debt. This holds as concerns existence and stability of the balanced growth
path. The stronger the response of the primary surplus to public debt, β, the sooner
the model is stable. Further, for certain parameter constellations the model converges to
stable limit cycles implying that the economy is characterized by sustained fluctuations.
3. As to growth effects of deficit financed increases in public investment, a deficit
financed increase in public investment reduces the balanced growth rate for large values

23
of β because the feedback effect of the increase in public debt outweighs the initial increase
in public investment. If β is small, an increase in public investment raises the balanced
growth rate but the model is unstable in this case.
4. Analyzing the transition path we could show that there is an overshooting of the
growth rates over the long-run balanced growth rate following a deficit financed increase
in public investment. The reason is that income effects matter as concerns the growth
rates on the transition path although the income effect is irrelevant for the determination
of the long-run balanced growth path.
A last remark refers to the public sector in our model. In our economy public invest-
ment is the only type of expenditure the government undertakes. Consequently, higher
public debt leading to an increase in the primary surplus can only reduce productive
public spending. But in reality other types of public spending, like unproductive public
consumption, could be reduced, too. However, looking at real world economies it indeed
seems that public investment is that type of expenditure which can be reduced most easily
as public debt rises. This holds because there is no obligation for governments to invest
in public infrastructure and there is no lobby group for public investment. Therefore,
the decline of public investment as a result of a rising public debt is not too surprising.
Empirical studies which support this view are for example Heinemann [16] who states
that public debt crowds out public investment or Gong et al. [11] who find this effect for
Germany and for the Netherlands.

References
[1] Arrow, Kenneth J., Kurz, Mordecai, 1970, Public Investment, the Rate of Return,
and Optimal Fiscal Policy. The John Hopkins Press, Baltimore.

[2] Aschauer, David A., 1989, ”Is Public Expenditure Productive?” Journal of Monetary
Economics, Vol. 23: 177-200.

24
 [3] Baier, Scott L., Glomm, Gerhard, 2001, ”Long-run growth and welfare effects of pub-
lic policies with distortionary taxation.” Journal of Economic Dynamics and Control,
Vol. 25: 2007-2042.

[4] Barro, Robert J., 1990, ”Government Spending in a Simple Model of Endogenous
Growth.” Journal of Political Economy, Vol. 98, no. 5, pt. 2: S103-25.

[5] Benhabib, Jess, Farmer, Roger, 1994, ”Indeterminacy and Increasing Returns.” Jour-
nal of Economic Theory, Vol. 63: 19-41.

[6] Blanchard, Olivier J., Fischer, Stanley, 1989, ”Lectures on Macroeconomics.” The
MIT Press, Cambridge, Massachusetts.

[7] Bohn, Henning, 1995, ”The sustainability of budget deficits in a stochastic economy.”
Journal of Money, Credit and Banking, Vol. 27, no. 1: 257-271.

[8] Bohn, Henning, 1998, ”The behaviour of U.S. public Debt and deficits.” Quarterly
Journal of Economics, Vol. 113: 949-963.

[9] Dhooge, Annick, Govaerts, Willy, Kuznetsov, Yuri A., Maestrom, W., Riet A.M.,
2003, ”CL MATCONT: A continuation toolbox in Matlab.” Ghent University,
http://allserv.rug.ac.be/∼ajdhooge/

[10] Futagami, Koichi, Morita, Yuichi, Shibata, Akihisa, 1993, ”Dynamic Analysis of
an Endogenous Growth Model with Public Capital.” Scandinavian Journal of Eco-
nomics, Vol. 95(4): 607-25.

[11] Gong, Gang, Greiner, Alfred, Semmler, Willi, 2001, ”Growth effects of fiscal policy
and debt austainability in the EU.” Empirica, Vol. 28, no. 1: 3-19.

[12] Glomm, Gerhard, Ravikumar B., 1997, ”Productive government expenditures and
long-run growth.” Journal of Economic Dynamics and Control, Vol. 21: 183-204.

25
[13] Greiner, Alfred, Hanusch, Horst, 1998, ”Growth and Welfare Effects of Fiscal Policy
in an Endogenous Growth Model with Public Investment.” International Tax and
Public Finance, Vol. 5, no. 3: 249-61.

[14] Greiner, Alfred, Koeller, Uwe, Semmler, Willi, 2004, ”Debt sustainability in the
European Monetary Union: Theory and empirical evidence for selected countries.”
CEM Working Paper, no. 71 (http://www.wiwi.uni-bielefeld.de/∼cem/).

[15] Hamilton, James D., Flavin M., 1986, “On the limitations of government borrowing:
A framework for empirical testing,” The American Economic Review, Vol. 76: 808-
819.

[16] Heinemann, Friedrich, 2002, ”Factor mobility, government debt and the decline in
public investment” ZEW Discussion Paper, No. 02-19, URL: ftp://ftp.zew.de/pub/
zew-docs/dp/dp0219.pdf

[17] Kremers, Jeroen M., 1988, ”US federal indebtedness and the conduct of fiscal policy,”
Journal of Monetary Economics, Vol. 23: 219-238.

[18] Khibnik, Alexander, Kuznetsov, Yuri, Levitin, Victor, Nikolaev, Eugene, 1993, ”Con-
tinuation techniques and interactive software for bifurcation analysis of ODE’s and
iterated maps,” Physica D 62: 360-371.

[19] Pfähler, Wilhelm, Hofmann, Ulrich, Bönte, Werner, 1996, ”Does Extra Public Cap-
ital Matter? An Appraisal of Empirical Literature.” Finanzarchiv N.F., Vol. 53:
68-112.

[20] Sturm, Jan E., Kuper, Gerhard H., de Haan, Jakob, 1998, ”Modelling Government
Investment and Economic Growth on a Macro Level.” in S. Brakman, H. van Ees and
S.K. Kuipers (eds.), Market Behaviour and Macroeconomic Modelling, pp. 359-406,
Mac Millan/St. Martin’s Press, London.

26
[21] Trehan, B. and C.E. Walsh, 1991, “Testing intertemporal budget constraints: theory
and applications to US Federal budget and current account deficits,” Journal of
Money, Credit and Banking, Vol. 23: 206-223.

[22] Turnovsky, Stephen, J., 1995, Methods of Macroeconomic Dynamics. MIT Press,
Cambridge, Mass.

[23] Turnovsky, Stephen, J., 2000a, ”The transitional dynamics of fiscal policy: Long-
run capital accumulation, and growth.” Department of Economics at the Univ. of
Washington, Seattle, Discussion Paper, no. 0018.

[24] Wilcox, D. W., 1989, “The sustainability of government deficits: Implications of the
present-value borrowing constraint,” Journal of Money, Credit and Banking, Vol. 21:
291-306.

27