Aschauer 2000

Public Capital and Economic Growth: Issues of
Quantity, Finance, and Efficiency
David Alan Aschauer

Bates College
I. Introduction
In the past decade, a large body of theoretical and empirical research has
considered the importance of the quantity of public capital required for
economic growth. For the most part, the empirical results of this line of
research point to a positive role for public capital in determining steady
state levels of output per capita and transitional growth rates. At the
same time, other work has pointed out the importance of the means of
financing government spending for economic growth. Here, the empiri-
cal results indicate a negative influence of higher government spending,
which acts as a proxy for a higher rate of taxation, on economic growth.
Finally, there is a budding literature on the importance of the effective-
ness, or efficiency, of public capital to the growth process. Here, the lim-
ited results in the literature suggest that an improvement in the efficiency
with which public capital is utilized stock has a meaningful, positive in-
fluence on growth.
This article develops a common framework to investigate the im-
portance of all the above mentioned aspects of the provision of public
capital for growth in output per worker. The following section fixes ideas
with a simple extension of the neoclassical growth model of R. M.
Solow and T. Swan.1 Next, I consider the relative importance of the three
aspects of public capital: how much you have, how you pay for it,
and how you use it. The final section presents my conclusions.
II. Conceptual Approach

The approach is an elaboration on the familiar neoclassical growth model
and only the essential elements are presented here. The analysis centers
on a Cobb-Douglas production function that relates output, Y, to various
sorts of capital, K, and labor, L. In this article, three types of capital will
be considered as inputs to the production process: private physical capi-
2000 by The University of Chicago. All rights reserved.

0013-0079/2000/4802-0009$02.00
This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
392 Economic Development and Cultural Change
tal, human capital, and public physical capital. Many recent articles have
indicated the separate importance of these types of capital for the growth
process.2 The production function, written in labor intensive form, is
n
yA
j1
k aj j , (1)
where y output per worker and k j type j capital per worker. It is

assumed that the production function exhibits constant returns to scale
across all inputs and, therefore, diminishing returns to all capital inputs,
so that j a j 1. Finally, A represents other, presently unspecified, fac-
tors that may be important to the production process.
In the steady state, there is an exogenous rate of technological prog-
ress, , and rate of growth of the labor force, . Each of the various
capital stocks is assumed to depreciate at the common rate . Conse-
quently, in the steady statewith unchanging capital stocks per effective
workerthe level of gross investment in each of the various types of
capital is given by
i j y ( ) k j , j 1, 2, . . . , n, (2)
where i j share of output devoted to gross investment in type j capital.

Substituting from equation (2) for the steady state capital stocks into
equation (1) and solving for y then yields the steady state level of output
per worker as

A i aj j

j 1
y() . (3)
( )
j
aj 1 j
aj
In the transition to the steady state, the growth rate of output per worker
is given by
ln

y(s)
y(0)
(1 exp(s)) ln
y()
y(0)
,
(4)
where represents the rate at which the economy converges to the

steady state. Substituting from equation (3) for the steady state level of
output in equation (4) yields
b ln ,
y(s) ij
ln c b 0 ln( y(0)) j (5)
y(0) j

David Alan Aschauer 393
where c is a constant and b 0 (1 exp(s)). The coefficients b j ,

j 1, 2, . . . , n, representing the effect of changes in the steady state
levels of type j capital on the transitional growth rate, are given by
aj b0
bj , j 1, 2, . . . , n. (6)
1
aj
j
This latter set of n equations can be solved for the output elasticities of
the various types of capital; specifically, we obtain
bj
aj , j 1, 2, . . . , n. (7)
j
bj b0
In the following empirical analysis, a stochastic version of equation (5)

will be estimated in order to obtain estimates of the convergence rate
[ ln(1 b 0)/s], growth sensitivities [b j, j 1, 2, . . . , n], and output
elasticities [a j , j 1, 2, . . . , n].
III. Data
The empirical analysis focuses on the importance of various kinds of
capital on the process of economic development. Consequently, the basic
data set covers 46 low- and middle-income countries over the period
197090. The definitions and sources of the data are:
y real gross domestic product per capita, with purchasing

power parity adjustment, from R. Summers and A.
Heston;3
i1 197090 average ratio of gross private investment to
output, from Summers and Heston;
i2 percentage of working age population in secondary
school, from Unesco yearbook;
i3 197090 average ratio of gross public investment to
output, from Summers and Heston;
average annual growth rate of population, from
Summers and Heston; and
rate of technological progress plus depreciation, assumed
to equal .05% per year.4

IV. Empirical Results

A. How Much You Have Is Important
Table 1 shows the results of estimating the basic model in three forms:
with private physical capital; with private physical capital and human
capital; and with private physical capital, human capital, and public
physical capital.
At first glance, the empirical results in equation (1) appear to be
broadly consistent with the predictions of the neoclassical growth model.
The steady state private physical capital stock (measured relative to out-
put) is highly positively correlated with output growth. A 10% increase
in the private capital stock ratio is shown to lead to a 3.2% increase in
output per capita over 2 decades, or roughly .16 per year. The implied
value of the output elasticity of private capital, .76, is high but consistent
with previous estimates in the literature.5 In addition, the negative coef-
ficient on the 1970 level of output per worker is consistent with a conver-
gence effect, whereby countries with relatively low levels of output per
capita grow at a relatively faster rate, though the implied convergence
rate is quite lowat .5 of 1% per yearand statistically quite weak.
However, it was precisely these kinds of results that led D. A.
Aschauer and N. G. Mankiw, D. Romer, and D. N. Weil, among others,
TABLE 1
Capital and Economic Growth in 46 Countries,
19701990
(Dependent Variable: ln( y(90)/y(70))
Explanatory Variable (1) (2) (3)
Constant .86 1.49 1.42

(.64) (.59) (.56)
ln( y(70)) .10 .36 .36
(.09) (.11) (.10)
Convergence rate .005 .022 .022
Private capital: .32 .24 .34
(.09) (.08) (.08)
Output elasticity .76 .28 .27
Human capital: .25 .24
(.07) (.07)
Output elasticity .29 .19
Public capital: .30
(.11)
Output elasticity .24
R2 .20 .37 .45
SER .37 .33 .30
CRTS test:
F-statistic 1.77 .01 .01
P-value .18 .98 .91
Note.Standard errors are in parentheses. SER

standard error of regression; CRTS constant returns to
scale.

to augment the basic Solow model by including a measure of human cap-

ital along with physical capital in the production function.6 Equation (2)
includes a proxy for human capitalthe secondary school enrollment
rate deflated by capitals effective depreciation rateand shows a clear
improvement in the explanatory power of the model. The private physi-
cal capital and human capital stock variables are statistically highly sig-
nificant and indicate an important quantitative role for both kinds of cap-
ital in growth. Specifically, 10% increases in physical capital and human
capital, respectively, are calculated to boost output by 2.4% and 2.5%
over 2 decades, or .12% per year. The coefficient on the 1970 level of
output per capita differs in a statistically significant manner from zero,
and the implied convergence rate, at 2.2% per year, is in the same range
as earlier estimates found in works by Mankiw, Romer, and Weil; R. J.
Barro and X. Sala-I-Martin; and elsewhere.7 The estimated convergence
rate is somewhat lower than the theoretical value
( ) (1 a 1 a 2), (8)
which, given the implied estimates of the output elasticities of private

physical capital and human capital, is equal to 3.3% per year.8 Finally,
the model as estimated contains the restriction that there are constant re-
turns to scale over the capital inputs included in the particular empirical
specificationhere, private physical capital and human capitaland la-
bor. This restriction is tested by running the regression equation in an
unrestricted formthat is, allowing the logarithm of the effective depre-
ciation rate to have a separate explanatory role in the equa-
tionand performing a Wald test. As indicated in the table, the data
contain virtually no evidence against this restriction; the relevant F-sta-
tistic takes on a very low value of less than .01 and carries a P-value of
.98.
In the economic growth literature, there is considerable controversy
regarding the relative importance of public and private physical capital
in the economic growth process. In a sample of 76 countries, Barro finds
that public capital investment and private capital investment have similar
effects on economic growth.9 In a sample composed of 100 countries
(a subset of which comprises the sample of 46 countries used in this
article), W. Easterly and S. Rebelo estimate an important role for infra-
structure capitalespecially transportation and communicationsin
economic growth.10 However, C. R. Hulten finds little effect of public
capital on economic growthafter controlling for the efficiency of use
of public capital.11
Equation (3) of table 1 includes the steady state measure of public
capital along with private capital and human capital. The coefficient esti-
mate on public capital is statistically and quantitatively important, and it
indicates that a 10% increase in public capital can be expected to raise

economic growth by 3.0% over 2 decades, or approximately .15% per

year. Generally speaking, the introduction of the public capital variable
leaves unaffected the estimated coefficients on the 1970 level of output
per capita and on human capital but raises the estimated coefficient on
private physical capital. The adjusted coefficient of determination rises
substantially, from .37 to .45, and a test of the constant returns to scale
restriction shows little evidence against the restricted model.
The Cobb-Douglas production structure allows a comparison of the
sample average marginal products of public and private physical capital
by use of the formula
mp 1
mp 3
a

i
1 3 ,
a3 i1
(9)
where mp 1 marginal product of private physical capital and mp 3

marginal product of public physical capital. Since the sample average
values of the ratios of private physical capital and public physical capital
to output are 1.37 and 1.32, respectively, we obtain
mp 1
1.09,
mp 3
which indicates that the data contain only weak evidence that a realloca-
tion of physical capital from public to private uses would exert a positive
influence on average growth.
B. How You Pay for It Is Important
A number of theoretical and empirical studies have pointed to the possi-
bility that a relatively large government sector places a burden on the
private sector and, thereby, may depress the rate of economic growth. In
an explicit optimizing framework, Barro shows how the benefits from
productive government spending need to be weighed against the costs
of distortional taxes that result in an optimal (i.e., growth maximiz-
ing) ratio of government spending to output, which is equal to the output
elasticity of government spending.12 Aschauer extends this analysis to
consider government capital and empirically estimates growth maximiz-
ing ratios of public capital to private capital.13
In the context of this study, I capture these notions in a tractable
fashion by postulating that the constant term in the production function
(1) now depends negatively on the tax burden associated with the accu-
mulation of public capital. The tax burden, in turn, is taken to be directly
related to the level of external public debt, expressed as a ratio to output,
which is issued (at least in part) to finance the initial acquisition of public
capital.14 Specifically, I assume
A A 0 exp(d Debt), (10)

where d 0 and Debt 1980 level of external public debt as a ratio to

output. This allows an expanded version of the growth equation of the
form
b ln d Debt.
y(s) ij
ln c b 0 ln(y(0)) j (11)
y(0) j
Table 2 presents estimates of the various specifications of the

growth model including the debt variable. As expected, in all three equa-
tions the public debt variable is negatively associated with growth in out-
put per capita, ranging from a low (absolute) value of .28 in equation (2)
to a high (absolute) value of .69 in equation (3). In the latter case, a 10%
increase in external public debt (i.e., by an amount equal to 3% of out-
put) can be expected to induce a 2.1% decrease in output over 2 decades,
or .1% per year.
In general, the other empirical results in table 2 conform closely to
those in table 1. In particular, all three types of capital are quantitatively
and statistically important in the determination of the rate of growth of
output per capita. As above, the addition of human capital (in eq. [2])
TABLE 2
Capital, Debt, and Economic Growth in 46
Countries, 19701990
Explanatory Variable (1) (2) (3)
Constant 1.21 1.69 1.85

(.66) (.61) (.51)
ln( y(70)) .13 .37 .36
(.09) (.11) (.09)
Convergence rate .007 .023 .022
Private capital .34 .27 .40
(.09) (.08) (.09)
Output elasticity .72 .31 .28
Human capital .23 .20
(.07) (.06)
Output elasticity .26 .14
Public capital .34
(.11)
Output elasticity .24
Debt .41 .28 .69
(.23) (.21) (.21)
R2 .24 .38 .56
SER .36 .32 .27
CRTS test:
F-statistic 1.25 .01 .25
P-value (.27) (.95) (.62)
Note.Standard errors are in parentheses. SER

standard error of regression; CRTS constant returns to
scale.

and public physical capital (in eq. [3]) results in a clear improvement in
the explanatory power of the model. The implied output elasticities of
the various forms of capital are reasonableparticularly in equation
(3)and the constant returns to scale restriction cannot be rejected at
conventional significance levels.
In the context of the empirical model, the growth maximizing ratio
of public capital to output is given by the expression
[ln(y(s)/y(0))] b3
d0 (12)
[i 3 /( )] i 3 /( )
or
i3 b
3. (13)
( ) d
Using the estimated coefficients for the growth sensitivities of public

capital (b 3 .34) and external debt (d .69), I find that the growth
maximizing level of public capital equals 49% of output. The actual sam-
ple average level of public capital equals 132% of outputso that fur-
ther increases in public capital, financed by external borrowing, can be
expected to diminish the economic growth rate of a representative coun-
try in the sample. Specifically, by differentiation of equation (11), hold-
ing fixed the 1970 level of output, I obtain
ln( y(90)) ln k 3 d Debt, (14)
so that with Debt k 3,
ln( y(90)) (b 3 d k 3) ln k 3. (15)
Thus, a 10% increase in public capital financed in this manner is esti-

mated to reduce output per capita by 6.0% over 20 years, or roughly .3%
per year. Consequently, despite the fact that public capital is beneficial
to growth in a gross sensewith an estimated output elasticity of .34
the average country in the sample appears to have accumulated an exces-
sive amount of public capital, resulting in a dampening effect on growth.

C. How You Use It Is Important

In his recent work, Hulten has presented empirical evidence that suggests
that the efficiency with which public capital is utilized is just asif not
moreimportant as the size of the public capital stock for the economic
growth process.15 Hulten defines the relationship between the effective
public capital stock, kg e, and the actual public capital stock, kg, as
kg e kg, (16)
where is a measure of the average level of public capital effectiveness.

In order to implement his model empirically, Hulten constructs a mea-
sure of public capital effectiveness by aggregating four indicators of pub-
lic capital performance (mainline faults per 100 telephone calls for tele-
communications, electricity generation losses as a percent of total system
output for power, and the percentage of paved roads in good condition
and diesel locomotive availability as a percent of the total rolling stock
for transportation) into an aggregate index. Noting that each of the indi-
vidual indicators is measured in its own units, Hulten sorts each of the
above indicators into quartiles, assigning values of .25, .50, .75, and
1.00, and then averages across quartile rankings for each performance
indicator to obtain an aggregate performance index.
In this article, I depart from Hultens approach in two directions.
First, an aggregate measure of public capital efficiency is derived from
the same basic data source but in a fashion that allows for a somewhat
more precise measure of efficiency. Specifically, each individual indica-
tor is normalized (as opposed to being given a quartile ranking) so that
performance is measured in terms of standard deviations from the aver-
age level of performance. The individual normalized indicators are then
averaged to obtain an aggregate performance indicator.
Second, for the sake of consistency with the normalized efficiency
measurewhich carries a mean value of zerothe average level of pub-
lic capital effectiveness is written as
exp( Eff ), (17)
where Eff is the public capital effectiveness measure. Here we note that
if Eff 0 then 1 and the capital stock is at an average level of
effectiveness. This allows the expanded growth equation
b ln
y(s) ij
ln c b 0 ln( y(0)) j
y(0) j
(18)
d Debt e Eff,
where the coefficient on the efficiency variable is given by e b 3 .

TABLE 3
Capital, Debt, Efficiency, and Economic Growth in 46
Countries, 19701990
Explanatory Variable (1) (2) (3) (4)
Constant 1.34 1.34 1.70 1.71

(.46) (.47) (.43) (.43)
ln(y(70)) .29 .31 .31 .31
(.08) (.09) (.08) (.07)
Convergence rate .017 .019 .019 .019
Private capital .17 .23 .26 .27
(.07) (.06) (.07) (.06)
Output elasticity .23 .27 .26 .27
Human capital .18 .19 .15 .15
(.06) (.06) (.05) (.05)
Public capital .11 .24 .28 .29
(.10) (.05) (.11) (.04)
Debt .56 .57
(.18) (.16)
Eff .33 .24 .29 .29
(.08) (.05) (.07) (.04)
R2 .62 .61 .69 .70
SER .25 .26 .23 .23
CRTS test:
F-statistic .01 .01 .10 .11
P-value .95 .97 .76 .75
Note.Standard errors are in parentheses. SER standard er-

ror of regression; CRTS constant returns to scale.
Table 3 presents results pertaining to the estimation of equation (18)

for the various definitions of capital. In this table, the first two equations
are estimated without the external public debt variable in order to allow
comparison with the results in Hulten. In equation (1) of table 3, the pub-
lic capital efficiency variable enters in a positive and statistically signifi-
cant manner. Quantitatively, a one-unit rise in the efficiency index can
be expected to induce a 3.3% increase in output per capita over 2 de-
cades, or about .16% per year. At the same time, the introduction of the
public capital efficiency variable erodes the statistical and quantitative
importance of the measured stock of public capital; the growth rate sen-
sitivity of the stock of public capital falls to .11 and is statistically weak-
ened, while the output elasticity of the stock of public capital drops to
.15.
These findings would seem to confirm the results in Hulten, which
led him to conclude that those countries that fail to use their infrastruc-
ture effectively pay a penalty in the form of lower growth rates and
that international aid programs aimed only at new infrastructure con-
struction may have a limited impact on economic growth, and may have

a perverse effect if they divert scarce domestic resources away from the
maintenance and operation of existing infrastructure stocks.16 However,
the potential importance of public capital stocks is enhanced in equation
(2) of table 3, which invokes the constraint that 1 so that the effec-
tive public capital stock is given by
kg e exp(Eff ) kg, (19)
which, in turn, implies an equality between the coefficients on the mea-

sured public capital stock and the efficiency variables. As is evident from
the results pertaining to equation (2) of table 3, the growth sensitivities
of the public capital stock and public capital effectiveness equal .24 and
are highly statistically significant. A test of the coefficient restriction
1 (or b 3 e) leads to a value of the relevant F-statistic equal to 2.17 and
an associated P-value of .15; consequently, the hypothesis of a parallel
importance of quantity and effectiveness of public capital cannot be re-
jected at conventional levels.
This argument for the importance of the quantity of public capital
is strengthened by the results in the third and fourth columns of table 3,
which include the external public debt variable to capture the adverse
effect of the financing of public capital on growth. In equation (3), the
coefficient on the public capital variable increases from .11 (the value of
the coefficient in eq. [1]) to .28 and becomes highly statistically signifi-
cant. In equation (4), which invokes the constraint that the quantity and
efficiency of public capital have parallel effects on growth, the coeffi-
cient on public capital is equal to .29 and nearly five times as large as
the associated standard error. These results point out in bold terms the
importance of considering both the level (and effectiveness) of public
capital and the means of financing public capital in order to properly as-
sess the effect of public capital accumulation on the growth process. Spe-
cifically, in the present sample of countries, the public capital measure
(based on public investment) and the external public debt variable them-
selves are positively correlated.17 While the former has a positive effect
on growth, the latter has a negative effect on growth so that the exclusion
of either variable from the regression equation can be expected to gener-
ate biased estimates of the effects of public capital and public debt on
growth.
D. Robustness Checks
A number of researchers have noted that estimated economic growth
equations using cross-country growth data sets may fail to be robust to
changes in model specification. For instance, Sala-I-Martin has looked
for stable relationships between economic growth and 59 explanatory
variables by running nearly 2 million regressions. He finds that 22 out
of the 59 variables appear to be significant.18 These include regional

variables, political variables, religious variables, market distortions,

types of investment, primary-sector production, openness, and type of
economic organization. It would be useful, then, to determine if the em-
pirical results shown in tables 13 are robust to the inclusion of some of
these variables.
Table 4 shows the results of including various sets of alternative
regressors, including regional dummies (present in all equations), a debt
crisis variable (Crisis), an openness variable (Openness), and a political
rights variable (Rights). The debt crisis variable is included in order to
control for the separate negative effect of the 1980s developing country
debt crisis on the economic performance of highly indebted countries
such as Mexico, Argentina, Nigeria, and Bolivia. This debt crisis vari-
able is constructed as a dummy variable, with a value of one for the
Baker 15, the group of countries where the gap between debt service
owed and ability to pay was the greatest in 1985 and for which Secretary
of the U.S. Treasury James Baker devised an assistance plan.19 The open-
ness variable is formed as the 1980 ratio of merchandise exports plus
imports to gross domestic product. The rights variable is constructed as
the sum of separate indicies of political rights and civil liberties for 1980
taken from R. D. Gastils work on world freedom; this variable is scaled
in such a way that higher values denote lower values of political rights
and civil liberty.20
Equations (1) and (2) of table 4 allow for separate intercept terms
for each of the following continental regions: Africa, Asia, South
America, and North America. The inclusion of separate growth effects
for each region has little effect on the overall explanatory power of the
regression and, more important, has no statistically significant effect on
the estimated coefficients of the model. There is, however, a slight reduc-
tion in the statistical significance of the human capital variable. Specifi-
cally, the standard error associated with the estimated coefficient on this
variable rises from .05 to .06, which lowers the associated t-statistic from
approximately 3 to 2.
It is quite conceivable that the negative influence of external public
debtas captured in the estimated equations in tables 2 and 3merely
reflects the effects of the 1980s debt crisis on economic growth rather
than the postulated adverse effect of tax rates on economic activity. That
is, the external public debt variable may simply be picking out those
countries most affected by the debt crisis.21 To consider this possibility,
equations (3) and (4) of the table include the debt crisis variable. Evi-
dently, the debt crisis had a quantitatively and statistically important ef-
fect on the average growth rates of the most heavily indebted countries
in the sample. Specifically, these latter countries had an average annual
rate of growth that fell .75% below that of the reference country in the
sample.22 However, the inclusion of the debt crisis variable has a rela-
tively insignificant effect on the estimated coefficients of the model. It

TABLE 4
Robustness in 46 Countries, 19701990
Explanatory Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Constant 1.80 1.77 1.39 1.41 1.49 1.50 1.47 1.49 1.58 1.59
(.60) (.59) (.46) (.45) (.62) (.60) (.63) (.61) (.63) (.61)
ln(y(70)) .30 .31 .26 .26 .26 .26 .25 .26 .25 .26
(.09) (.09) (.08) (.08) (.09) (.09) (.09) (.09) (.09) (.09)
Convergence rate .019 .019 .015 .015 .015 .015 .014 .015 .014 .015
Private capital .25 .25 .26 .27 .25 .26 .25 .26 .22 .23
(.08) (.06) (.07) (.06) (.07) (.06) (.08) (.06) (.08) (.07)
Output elasticity .25 .25 .29 .28 .27 .28 .28 .29 .24 .25
Human capital .13 .14 .15 .15 .13 .13 .13 .13 .12 .12
(.06) (.06) (.05) (.05) (.06) (.06) (.06) (.06) (.06) (.06)
Public capital .30 .29 .24 .27 .26 .28 .26 .29 .27 .29
(.12) (.05) (.11) (.04) (.12) (.05) (.12) (.05) (.12) (.05)
403
Debt .55 .57 .57 .59 .55 .57 .55 .57 .50 .52
(.19) (.16) (.17) (.15) (.19) (.16) (.19) (.17) (.20) (.18)
Eff .29 .29 .28 .27 .30 .28 .30 .29 .30 .29
(.07) (.05) (.07) (.04) (.07) (.05) (.07) (.05) (.07) (.05)
Crisis .15 .15 .18 .18 .19 .18 .20 .19
(.09) (.09) (.11) (.11) (.11) (.11) (.11) (.11)
Openness .02 .02
(.06) (.06)
Rights .01 .01
(.02) (.02)
R2 .66 .67 .70 .71 .68 .69 .67 .68 .68 .68
SER .24 .24 .22 .22 .23 .23 .24 .23 .23 .23
CRTS test:
F-statistic .03 .03 .10 .12 .05 .06 .23 .25 .03 .03

P-value .85 .85 .75 .73 .82 .80 .63 .62 .86 .86
Note.Standard errors are in parentheses. SER standard error of regression; CRTS constant returns to scale.
appears, then, that the negative influence of the external public debt vari-
able on economic growth is reflecting more than the debt crisis and is at
least consistent with a negative tax effect.
Many authors have found a significant relationship between eco-
nomic growth and development and various indicators of the quality of
the overall economic environment.23 For example, a high degree of inte-
gration into the international economic system as well as a strong adher-
ence to political rights and civil liberties can be expected to be conducive
to positive economic performance. Accordingly, the last four equations
of table 4 include as separate regressors the openness variable (in eqq.
[7] and [8]) and the rights variable (in eqq. [9] and [10]). In my sample,
it appears that neither variable has much explanatory power, over and
above the remaining regressors (including the regional dummies). In ad-
dition, neither variable has much effect on the estimated coefficients of
the capital, debt, or efficiency variables. In particular, the output elastici-
ties for private capital, human capital, and public capital are calculated
to be in the range of .22.26, .12.13, and .26.29, respectively, in close
conformity to their values shown in table 3.
V. Conclusion
In this article I have extended the neoclassical model to assess the impor-
tance of three aspects of government intervention on economic growth
on the transition path to the steady state. First, public physical capital is
included along with private physical capital and human capital as an in-
put in the steady state production function. Second, the means of financ-
ing public capital is allowed to affect the level of productivity. Third,
the efficiency of use of public capitalalong with the quantity of public
capitalis taken to determine the effective public capital stock.
In this setting, three questions pertaining to economic growth may
be asked: Does how much public capital you have matter? Does how
you finance public capital matter? and, Does how you use public capital
matter? The empirical results presented in this article allow affirmative
answers to each of these questions. Specifically, 10% increases in either
the quantity or the efficiency of public capital are estimated to increase
output per capita by 2.9% over 2 decades while a 10% increase in exter-
nal public debt is estimated to decrease output per capita by 1.7% over
the same time frame. Thus, an average increase in public capital, fi-
nanced by external debt, is estimated to detract from economic growth
in an amount equal to some .25% per year while an above average
increase in public capitaldefined as a simultaneous increase in quantity
and efficiency of public capitalis estimated to have a minor positive
effect on economic growth by an amount equal to .1% per year. The
main lesson to be drawn from these findings is that in formulating eco-
nomic development policies, countries are well advised to pay as much

attention to how public capital is financed and used as to how much pub-
lic capital is accumulated.
Notes
1. R. M. Solow, A Contribution to the Theory of Economic Growth,
Quarterly Journal of Economics 70 (February 1956): 6594; T. Swan, Eco-
nomic Growth and Capital Accumulation, Economic Record 32 (October
1956): 33461.
2. For the importance of human capital, see D. A. Aschauer, Is Public
Education Productive? in Higher Education and Economic Growth, ed. W. E.
Becker and D. R. Lewis (Norwell, Mass.: Kluwer Academic Press, 1991); N. G.
Mankiw, D. Romer, and D. N. Weil, A Contribution to the Empirics of Eco-
nomic Growth, Quarterly Journal of Economics 107 (May 1992): 40737. For
the importance of public capital, see D. A. Aschauer, How Big Should the
Public Capital Stock Be? The Relationship between Public Capital and Eco-
nomic Growth, Public Policy Brief no. 43 ( Jerome Levy Economics Institute,
Annandale-on-Hudson, N.Y., 1998).
3. R. Summers and A. Heston, The Penn World Table (Mark 5): An Ex-
panded Set of International Comparisons, 19501988, Quarterly Journal of
Economics 106 (May 1991): 32768.
4. The empirical results are not sensitive to other reasonable assumed val-
ues for technological progress and depreciation.
5. See, for instance, the results for the textbook Solow model in the fol-
lowing articles: Mankiw, Romer, and Weil; and W. Nonneman and P. Vanhoudt,
A Further Augmentation of the Solow Growth Model and the Empirics of Eco-
nomic Growth for OECD Countries, Quarterly Journal of Economics 111 (Au-
gust 1996): 94353.
6. See D. A. Aschauer, Is Public Education Productive? and Mankiw,
Romer, and Weil.
7. R. J. Barro and X. Sala-I-Martin, Convergence across States and Re-
gions, Brookings Papers on Economic Activity, no. 1 (1991): 10782.
8. This theoretical value pertains to a closed economy version of the neo-
classical growth model. The convergence rate can be expected to be higher in
an open economy version of the neoclassical growth model that allows (partial)
capital mobility, as demonstrated in R. J. Barro, N. G. Mankiw, and X. Sala-I-
Martin, Capital Mobility in Neoclassical Models of Growth, American Eco-
nomic Review 85 ( January 1995): 10315. The countries in the data sample that
I used in this article are, for the most part, net debtors in the international capital
market. Thus, an estimated convergence rate that lies somewhat below the theo-
retical value for a closed economy version of the model presents something of
an empirical puzzle.
9. R. J. Barro, Economic Growth in a Cross Section of Countries, Quar-
terly Journal of Economics 106 (May 1991): 40743.
10. W. Easterly and S. Rebelo, Fiscal Policy and Economic Growth,
Journal of Monetary Economics 32 (December 1993): 41758.
11. C. R. Hulten, Infrastructure Capital and Economic Growth: How
Well You Use It May Be More Important than How Much You Have, Working
Paper no. 5847 (National Bureau of Economic Research, Cambridge, Mass., De-
cember 1996).
12. R. J. Barro, Government Spending in a Simple Model of Endogenous
Growth, Journal of Political Economy 98 (October 1990): S103S125.
13. D. A. Aschauer, Do States Optimize? Public Capital and Economic

Growth, Working Paper no. 189 (Jerome Levy Economics Institute, April
1997).
14. Strictly speaking, the total tax burden associated with a certain level of
public capital can be expressed in the following way. In the steady state, the
government must raise tax revenues equal to (a) the interest charge associated
with the initial purchase of government capital and (b) the ongoing gross invest-
ment necessary to maintain the public capital stock against technological prog-
ress, population growth, and physical depreciation. Letting k 3 represent public
capital, r the real interest rate, and a tax rate on labor and capital income,
y r k 3 ( ) k 3 (r ) k 3 .
Assuming that public debt is used to finance the initial acquisition of public capi-
tal, we have
(r ) Debt,
where Debt denotes the ratio of public debt to output. Thus, the tax burden is
associated with the ratio of public debt to output. In the empirical work, external
public debt is used as a proxy for total public debt, since data on total public
debt are unavailable for many of the countries in the sample. Also, the empirical
results are not particularly sensitive to the use of Debt or as the measure of
the tax burden; accordingly tables 2 and 3 only report results from empirical
equations using Debt.
15. See Hulten.
16. Ibid., pp. 23 and 25.
17. Specifically, the simple correlation coefficient between the public capi-
tal and external public debt variables equals .39.
18. See X. Sala-I-Martin, I Just Ran Two Million Regressions, Ameri-
can Economic Review Papers and Proceedings (May 1997): 17883, quote on
179.
19. See J. Cavanagh, F. Cheru, C. Collins, et. al, From Debt to Develop-
ment (Washington, D.C.: Institute for Policy Studies, 1985). The Baker 15
countries, ranked in terms of their external debt (high to low), are Brazil, Mex-
ico, Argentina, Venezuela, Phillipines, Chile, Yugoslavia, Nigeria, Morocco,
Peru, Columbia, Ecuador, Ivory Coast, Uruguay, and Bolivia.
20. R. D. Gastil, Freedom in the World (Westport, Conn.: Greenwood,
1981).
21. I thank an anonymous referee for this suggestion.
22. The reference country is the sole European country in the sample,
namely, Portugal.
23. See Sala-I-Martin.


Aschauer 2000

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Aschauer 2000

Uploaded by

Copyright:

Available Formats

Public Capital and Economic Growth: Issues of

Quantity, Finance, and Efficiency

David Alan Aschauer

II. Conceptual Approach

2000 by The University of Chicago. All rights reserved.

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

where y output per worker and k j type j capital per worker. It is

where i j share of output devoted to gross investment in type j capital.

where represents the rate at which the economy converges to the

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

where c is a constant and b 0 (1 exp(s)). The coefficients b j ,

In the following empirical analysis, a stochastic version of equation (5)

y real gross domestic product per capita, with purchasing

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

IV. Empirical Results

Explanatory Variable (1) (2) (3)

Constant .86 1.49 1.42

Note.Standard errors are in parentheses. SER

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

to augment the basic Solow model by including a measure of human cap-

which, given the implied estimates of the output elasticities of private

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

economic growth by 3.0% over 2 decades, or approximately .15% per

where mp 1 marginal product of private physical capital and mp 3

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

where d 0 and Debt 1980 level of external public debt as a ratio to

Table 2 presents estimates of the various specifications of the

Explanatory Variable (1) (2) (3)

Constant 1.21 1.69 1.85

Note.Standard errors are in parentheses. SER

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

Using the estimated coefficients for the growth sensitivities of public

ln( y(90)) ln k 3 d Debt, (14)

so that with Debt k 3,

ln( y(90)) (b 3 d k 3) ln k 3. (15)

Thus, a 10% increase in public capital financed in this manner is esti-

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

C. How You Use It Is Important

where is a measure of the average level of public capital effectiveness.

exp( Eff ), (17)

where the coefficient on the efficiency variable is given by e b 3 .

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

Explanatory Variable (1) (2) (3) (4)

Constant 1.34 1.34 1.70 1.71

Note.Standard errors are in parentheses. SER standard er-

Table 3 presents results pertaining to the estimation of equation (18)

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

kg e exp(Eff ) kg, (19)

which, in turn, implies an equality between the coefficients on the mea-

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

variables, political variables, religious variables, market distortions,

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

This content downloaded from 128.252.067.066 on July 25, 2016 20:16:37 PM

You might also like