The Impact of Government Size on Economic Growth:

A Time Series Cross-Country Study

Atul Dar and Saleh Amirkhalkhali*

Introduction

There is a vast empirical literature on economic growth, its patterns and

determinants. Over the past 10–15 years or so, a large portion of that literature
has focused, within an aggregate growth-accounting framework, on the role of
factors such as trade orientation in affecting the growth performance of an
economy. More recently, the literature has sought to incorporate new theoretical
developments (such as endogenous growth models) and to test their
implications — see, for instance, Jones (1995) and Mankiw et al.(1992). On the
other hand, there is a political economy literature which argues that ‘economics
alone cannot fully explain the enormous variance across countries in growth
and, more generally, in economic outcomes and policy choices’ (Alesina and
Perotti, 1994: 351). This approach stresses the importance of institutions,
political stability and inequality of income in determining growth via their
influence on the nature and quality of government policies. At the heart of these
developments is the debate about the causes of differences in economic growth
rates among nations.
A major thrust of the new growth theory is that government policy (via its
impact on the rate of technological change, for instance) can permanently
change a country’s long-term rate of economic growth. In this article we
examine the role played by the size of the government sector in modifying the
impact of various determinants on economic growth. This is attempted by
extending our previous study (Amirkhalkhali and Dar, 1995) of the impact of
openness and country size on economic growth. Our focus is on a sample of 17
Western hemisphere countries for which we have data from the early 1960s to
the mid-1990s.
The article is organised as follows. The following section contains a brief
discussion on economic growth and government variables, followed by a
section discussing the data, models and estimation techniques. The fourth

* Professor of Economics and Professor and Chairperson, Economics Department, Saint Mary’s
University, Halifax, NS, Canada, respectively.

Development Policy Review Vol. 17 (1999), 65–76

© Overseas Development Institute 1999. Published by Blackwell Publishers, 108 Cowley Road,
Oxford OX4 1JF, UK, and 350 Main Street, Malden, MA 02148, USA.
66 Development Policy Review

section discusses our empirical findings, and we conclude with a summary of

the findings and some conclusions drawn from them.

Economic growth and government size

There is a substantial theoretical as well as empirical literature on the

relationship between economic growth and government variables (see Tanzi and
Zee (1997) for a compact survey of the issues). The three main fiscal
instruments — taxation, expenditure and the aggregate budgetary
balance — impact on long-term growth via their effects on the efficiency of
resource use (given the technology), the rate of factor accumulation, and the
pace of technological advance. These links can be complex, a factor
compounded by the interdependence between the instruments themselves.
Empirical studies of the relationship between economic growth and tax and/or
public expenditure variables (expressed either in terms of rates of change or as
a percentage of GDP) face the significant problem not only of isolating their
independent impacts accurately, but also of capturing the effects that the
structure of taxation and government spending has on growth, effects which are
likely to be at least as important, if not more so, than those associated with
aggregate measures of taxation and public spending.
In the light of these difficulties, our approach to the role of policy in
determining growth rates adopts a somewhat different approach. Rather than
focusing on the impacts of specific government policy instruments, such as
taxation and public expenditure, we consider whether the overall size of the
government sector has implications for economic growth. The overall size of
government is measured by government expenditure plus tax revenue relative
to GDP. Theoretically speaking, there is uncertainty about the direction of the
impact of the size of government (measured in some way) on the rate of growth.
This is because there are persuasive arguments for both negative impacts
(through government inefficiency, the excess burden of taxation, etc.) and
positive impacts (through beneficial externalities through the development of
a legal, administrative, and economic infrastructure, and through interventions
to offset market failures, etc.) on growth (see, for instance, Ram, 1986 and
Tanzi and Zee, 1997). At any rate, whatever the direction of the relationship
between government size and growth performance, it is not clear whether that
relationship is monotonic. The empirical literature on the issue is relatively
sparse and the results can be sensitive to the way the size of government is
measured (see Rubinson, 1977 and Landau, 1983).
In our study of economic growth in 17 Western hemisphere countries, we
adopt the basic growth accounting, production function model of Solow (1956)
in which the rate of economic growth is a function of capital and labour
accumulation and total factor productivity. We explicitly assume that total factor
Dar and Amirkhalkhali, Impact of Government Size on Economic Growth 67

productivity depends, in turn, upon the rate of export expansion. Rapid export
growth implies a dynamic export sector, most likely characterised by high rates
of absorption of new technologies from abroad as well as more rapid
technological progress at home. To assess how the impact of these factors on
growth is influenced by the size of government, we classify countries according
to the aggregate size of government in the economy, and study the relationship
specified above for each category of government size using the Swamy-Mehta
random coefficients approach. In other words, our interest is in determining
whether, and to what extent, the impacts on economic growth of traditional
factors such as capital accumulation, growth of the labour force, and export
expansion are affected by the overall size of government in the economy. The
random coefficients approach is designed to accommodate inter-country
differences in the nature and structure of institutions and other forms of
heterogeneity, all of which are central to the political economy literature, but
which are very difficult to quantify in unique and reliable fashion. Our approach
would, therefore, probably be superior to those that attempt to quantify these
factors directly.

The data, models and estimation strategy

In this article, we use time series data for the following 17 Western hemisphere
countries: Argentina (1964–95), Bolivia (1967–96), Brazil (1965–95), Canada
(1962–96), Chile (1963–96), Colombia (1970–96), Costa Rica (1962–96), El
Salvador (1962–96), Guatemala (1962–96), Honduras (1962–96), Mexico
(1962–96), Nicaragua (1962–96), Panama (1961–95), Peru (1963–95), the
United States (1961–96), Uruguay (1962–96), and Venezuela (1961–96). The
data were obtained from various issues of Government Finance Statistics and
International Financial Statistics Yearbooks published by the International
Monetary Fund.
Table 1 presents sample period averages for rates of growth of output (G Y),
labour (GL), and exports (GX), as well as the investment to GDP (denoted by
I/Y) and total exports (X) plus imports (M) to GDP ratios, and population
(1995–6) for the countries in the sample. The countries have been classified into
four groups in terms of the size of government measured as the ratio of
government expenditure (G) plus revenue (R) to GDP. It can be seen that the
size of government varies from a relatively low 21-26% for Guatemala, Bolivia,
Colombia and El Salvador, to a high in the 49–58% range for Brazil, Uruguay,
Venezuela, Nicaragua, Chile and Panama. Argentina, Peru and Honduras are
in the intermediate group of 30–31% range, while Mexico, Canada, the US and
Costa Rica belong to the intermediate-to-high group III, with the size of
government ranging from about 36% to 44%. The countries with the lowest
(Nicaragua) and highest (Panama) rates of economic growth both belong to
68 Development Policy Review

group IV - the group characterised by the biggest size of government. However,

apart from this, growth rates, in general, appear to be quite evenly distributed
across the groups.

Table 1
Average annual growth of output (GY) and exports (GX), investment-
output ratios (I/Y), openness ((X+M)/Y), government finance ((G+R)/Y)
and population

Country Period GY GY I/Y (G+R)/Y (X+M)/Y Population

(million)
I. Guatemala 1962-96 4.0 9.7 14.1 20.6 41.5 11
(0.72) (1.64) (0.22) (0.14) (0.11)
Bolivia 1967-95 3.3 9.9 16.6 24.1 46.8 8
(1.03) (2.55) (0.22) (0.24) (0.07)
Colombia 1970-96 4.5 12.2 19.6 25.3 30.2 36
(0.43) (1.25) (0.11) (0.08) (0.19)
El Salvador 1962-96 3.4 7.7 15.8 26.0 56.3 6
(1.28) (2.17) (0.23) (0.15) (0.18)
II. Argentina 1964-95 2.4 10.5 20.1 30.1 17.9 35
(2.16) (1.86) (0.20) (0.21) (0.32)
Peru 1963-95 3.0 8.6 19.9 30.4 34.2 24
(1.85) (2.15) (0.26) (0.15) (0.19)
Honduras 1962-96 4.0 8.7 21.0 31.0 67.5 6
(0.84) (1.37) (0.29) (0.08) (0.17)
III. Mexico 1962-96 4.7 16.5 21.6 35.4 23.3 97
(0.82) (1.41) (0.13) (0.23) (0.27)
Canada 1962-96 3.8 10.8 22.2 41.8 50.3 29
(0.65) (0.75) (0.12) (0.06) (0.18)
USA 1961-96 2.9 10.6 18.4 43.1 17.1 267
(0.74) (1.02) (0.08) (0.04) (0.30)
Costa Rica 1962-96 4.8 11.3 23.6 44.0 67.5 3
(0.74) (0.95) (0.14) (0.14) (0.14)
IV. Brazil 1965-95 5.1 12.6 22.9 49.3 16.5 158
(0.93) (1.22) (0.17) (0.19) (0.21)
Uruguay 1962-96 2.1 8.6 13.2 50.0 37.7 3
(1.94) (1.66) (0.14) (0.14) (0.22)
Venezuela 1961-96 3.6 10.7 23.3 50.1 52.2 22
(3.00) (4.01) (0.33) (0.23) (0.14)
Nicaragua 1962-96 2.2 8.4 20.6 56.4 60.3 5
(3.39) (2.41) (0.24) (0.44) (0.19)
Chile 1963-96 4.1 12.9 18.8 56.8 48.7 15
(1.38) (1.94) (0.33) (0.26) (0.31)
Panama 1961-95 5.4 10.5 23.2 57.6 80.9 3
(0.96) (1.55) (0.30) (0.09) (0.14)

Note: Figures in parentheses are coefficients of variation.

Dar and Amirkhalkhali, Impact of Government Size on Economic Growth 69

If we measure openness in terms of the ratio of total trade relative to GDP,

Table 1 shows that Argentina and Mexico fall into the group of least open
countries, while Panama, Honduras, Costa Rica and Nicaragua are the most
open economies. Countries such as Canada and Chile are in the intermediate
range in terms of this definition of openness. It is interesting to note that the
degree of openness appears to bear no systematic relation to the size of
government involvement in the economy — in particular, countries with more
government are not necessarily the least open. Rather, country size (as measured
by population) appears to be more systematically related to the degree of
openness; smaller countries tend to trade proportionately more than larger
countries, other things being equal. However, variations in openness reflect
differences not only in size, but also in economic policy and location. For
instance, the sheer size of the United States puts it at the bottom of the openness
list, but ahead of Brazil, perhaps as a result of its more liberal economic policy
framework. Argentina is not very open, not so much because of its size, but
because of the pervasive controls on trade that have historically characterised
its economic policy. Uruguay, with the smallest population in the sample, ranks
low in terms of openness, again probably not because of location or size but
rather because of the policy framework. By the same token, the intermediate-to-
high openness of Venezuela and Canada, in spite of their relatively large size,
may well reflect not only their relatively liberal economic policy framework but
also their favourable location.
The model employed in this article is a generalisation of the commonly used
growth accounting model based on the concept of an aggregate production
function. In this approach, export growth (via its favourable impact on the
efficiency of resource use, innovative activity and the rate of technical progress,
and the realisation of economies of scale) raises total factor productivity growth
and, by implication, economic growth. The production function approach to
export expansion and economic growth involves the following aggregate
production function:

Y = Y(K, L, X) (1)

where Y is aggregate real output, K is the capital input, L is the labour input and
X represents exports. Assuming that the aggregate production function is linear
in natural logarithms, the usual model employed for empirical purposes (after
taking total derivatives and manipulating) can be shown to be:

(GY)it = 1 + 2(I/Y)it + 3(GX)it + 4(GL)it + uit (2)

where GY is the rate of growth of Y, I/Y the investment-output ratio, GX the rate
of growth of X, and GL the rate of growth of L, approximated, here, by the rate
of population growth. Note that 2 is the marginal physical product of capital,
70 Development Policy Review

3 and 4 are the partial elasticities of Y with respect to X and L respectively,

and u is the disturbance term. The subscripts i (i=1,2,...,N) and t (t=1,2,...,T)
index the countries and time periods in the sample respectively.
In recent years, this model has become a popular one for studying the
determinants of the rate of economic growth (see, for instance, Ram, 1985;
Alam, 1991). A major difficulty with these studies is that they all employ, as the
underlying analytical tool, a fixed coefficients version of the production
function. However, this is a major weakness of the model in that it assumes
away inter-country differences by virtue of the assumption that all coefficients
are the same across countries. This is a questionable assumption a priori, and
one that should at least be treated as a testable proposition. We overcome this
problem by adopting the more general random coefficients model which permits
us to treat the fixed coefficients assumption as a testable proposition. In
addition, the random coefficients model can be seen as a refinement of the
stochastic law relating economic growth to its main determinants (see Pratt and
Schlaifer, 1988).
In dealing with these problems, we first postulate :

(GY)it = 1 + 2(I/Y)it + 3(GX)it + 4(GL)it + WitU (3)

where W is the set of excluded variables that, along with those that are included,
is sufficient to determine GY. However, in the linear deterministic law stated
by equation (3), neither the slope coefficients nor W are unique in that they are
sensitive to the parameterisation chosen. To ensure uniqueness, we postulate:

Wit = 1i + 2i +(I/Y)it + 3i (GX)it + 4i (GL)it + vit (4)

Substituting equation (3) for (2) yields:

(GY)it = 1i + 2i (I/Y)it + 3i (GX)it + 4i (GL)it + uit (5)

where 1i = 1 + 1iU , 2i = 2 + 2iU , 3i = 3 + 3iU , 4i = 4 + 4iU and uit =
vitU .
Note that equation (5) is a random coefficients model, and that the
disturbance is not the joint effect of the excluded variables; instead, it is the
joint effect of the remainder of the excluded variables after the effect of the
included variables has been factored out. Note also that, whereas the included
variables cannot be uncorrelated with every variable that affects GY, they can
be uncorrelated with the remainder of every such variable (see Pratt and
Schlaifer, 1988). Thus, each of our explanatory variables can be uncorrelated
with u, and equation (5) can be taken to represent the law relating GY to its
determinants.
Dar and Amirkhalkhali, Impact of Government Size on Economic Growth 71

The random coefficients model represented by equation (5) is more general

than models employed in previous studies, not only because it describes more
correctly the law relating GY to its determinants, but also because it
accommodates inter-country heterogeneity.

The empirical results

Table 2 reports the OLS results for each group as well as the entire sample, and
Table 3 gives the corresponding GLS results. Looking at the OLS results, we
note that, with the exception of group II, each group regression as well as the
regression using the entire sample is highly significant as a whole. Most
coefficients are also significant at the 5% level. The OLS results, however, are
not reliable if the impacts on growth are country-specific. Furthermore, the
Durbin-Watson (DW) statistics and the Breusch-Pagan-Godfrey (BPG) tests
reported in Table 2 point to the existence of autocorrelation in all of them and
heteroscedasticity for groups I and IV. We turn therefore to our random
coefficients GLS results.

Table 2
Regression results: OLS
Model: (GY)it = 1 + 2(I/Y)it + 3(GX)it + 4(GL)it + uit

2 3 4 1 F R2 DW BPG No Nc

GROUP I 0.252* 0.066* 0.844* -3.109* 15.9* 0.28 0.92 9.28* 125 4
(0.071) (0.014) (0.343) (1.565)

GROUP II 0.009 0.046** 0.965** 0.140 1.8 0.05 1.53 3.23 101 3
(0.092) (0.024) (0.504) (2.303)

GROUP III 0.097** 0.055* 1.172* 0.973 13.0* 0.22 1.25 1.11 141 4
(0.066) (0.014) (0.282) (1.583)

GROUP IV 0.308* 0.046* -0.403 2.122** 14.6* 0.18 1.32 19.6* 205 6
(0.059) (0.014) (0.393) (1.151)

ALL 0.173* 0.053* 0.511* -1.439* 29.8* 0.14 1.27 10.4* 572 17
(0.032) (0.009) (0.190) (0.714)

Note: Figures in parentheses are the estimated standard errors. No and Nc denote the number of
observations and number of countries respectively. F gives the F-ratio for testing the overall
significance of the model. DW and BPG are the Durbin-Watson and Breusch-Pagan-Godfrey
statistics respectively. * indicates statistical significance at the 5% level.
72 Development Policy Review

Table 3 contains the aggregate group results based on GLS estimation. Note
first that the G-statistic is statistically significant at the 5% level in all cases,
thereby vindicating the random coefficients model. What do the results suggest
about the size of government ? First, it is evident that investment rates have, on
average, the largest statistically significant impact on economic growth for
group IV, the group of countries which are characterised by the largest size of
government. The corresponding impact for group I is somewhat smaller but also
significant. Interestingly, this impact for the intermediate groups is not
significant. Since the investment rate, on average, is roughly similar across
groups, these absolute impacts are also suggestive of their relative impacts. It
seems, therefore, that the role of capital formation in fostering economic growth
is not adversely affected for countries with big government sectors. This may
well reflect the beneficial externality that government activity generates in the
economy. The results for the impact of export growth are also more clear-cut
here. The coefficients are significant at the 5% level for all groups taken
together, and although there does not appear to be much difference in the
absolute growth impacts for the four groups there is a positive (at least non-
decreasing) relationship between the size of the impact and the size of the
government sector. However, the impact is statistically significant only for the
countries where government size is largest (group IV) and smallest (group I).

Table 3
Regression results: random coefficients GLS
Model:(G<)it = 1i + 2i (I/Y)it + 3i (G;)it + 4i (G/)it + uit

2 3 4 1 G-statistic No Nc

GROUP I 0.230* 0.056* 0.599 -2.096 40.2* 125 4

(0.092) (0.023) (0.992) (2.559)

GROUP II 0.023 0.058 0.732 0.020 21.3* 101 3

(0.069) (0.055) (0.686) (2.411)

GROUP III 0.320 0.064 1.317 -6.029 45.7* 141 4

(0.214) (0.046) (0.920) (4.934)

GROUP IV 0.270* 0.071* 0.693 -3.626 46.9* 205 6

(0.111) (0.029) (1.429) (3.402)

ALL 0.206* 0.068* 0.806 -2.627 174.0* 572 17

(0.073) (0.018) (0.585) (1.879)

See notes to Table 2.

It is more instructive, particularly since the G-statistic supports the random

coefficients model, to look at the country-specific results, which may display
Dar and Amirkhalkhali, Impact of Government Size on Economic Growth 73

Table 4
Regression results: random coefficients GLS
Model: (GY)it = 1i + 2i (I/Y)it + 3i (GX)it + 4i (GL)it + uit

Group/Country 2 3 4 1
I. Guatemala 0.1368 0.1065 0.3515 -0.0443
0.0941 0.0213 0.7870 2.6668
Bolivia 0.3679 0.0396 -1.3960 -0.1774
0.0954 0.0160 0.4771 1.9601
Colombia 0.1531 0.0568 1.3487 -2.2096
0.0980 0.0190 0.6435 2.4780
El Salvador 0.2365 0.0359 2.4887 -6.9467
0.1144 0.0281 0.6071 2.4089
II. Honduras -0.0308 0.1568 -0.1054 3.6239
0.662 0.0266 0.6588 2.1093
Argentina 0.1573 0.0325 1.2520 -3.0433
0.1174 0.0277 0.7990 2.6716
Peru 0.0469 0.0906 1.3062 -1.5539
0.0576 0.0199 0.4805 1.6588
III. Mexico 0.2628 0.0240 3.1346 -9.6990
0.1063 0.0197 0.6194 2.3973
Canada 0.3238 0.0822 0.3078 -4.6112
0.0970 0.0276 0.5673 2.3029
USA 0.5131 0.0024 0.9889 -7.6597
0.1216 0.0239 0.5022 2.2759
Costa Rica -0.0721 0.1450 0.8346 2.5632
0.0842 0.0284 0.7714 2.5919
IV. Brazil 0.2478 0.0877 1.6704 -5.4992
0.1021 0.0297 0.8662 2.7071
Uruguay 0.1297 0.1027 0.4062 -0.7531
0.1160 0.0253 0.5241 1.5659
Venezuela 0.1463 0.0045 1.3012 -3.5799
0.0809 0.0151 0.8685 2.0732
Nicaragua 0.3396 0.1300 -2.4915 2.2504
0.1189 0.0302 0.8250 2.6379
Chile 0.4156 0.0365 1.4407 -6.9905
0.1023 0.0243 0.8092 2.5334
Panama 0.1272 0.0750 1.4831 -2.4265
0.0898 0.0299 0.8605 2.5001

variations that are obscured in the aggregate results. These results are presented
in Table 4 . It can be seen that the investment impact is the largest for the US.
Other countries with large impacts are Chile, Bolivia, Nicaragua, Canada and
Mexico. Interestingly, these countries straddle all four groups, with three of
them belonging to the intermediate group. It is evident that the aggregate group
74 Development Policy Review

results mask some important differences between countries. Thus, in Table 3

group III countries showed the largest but statistically insignificant growth
impact of investment. However, it seems that this is probably because the group
III aggregate impacts are pulled down by the negative impact for Costa Rica
(especially in light of the fact that other countries in group III are, as already
noted, among those with the largest impacts in the entire sample). At the same
time, it is clear that the countries in group IV are also those with large (and
statistically significant) impacts. The country-specific results thus indicate that
the relationship between the size of the investment impacts on growth and the
size of government is not monotonic. A fixed coefficients approach would
therefore give a misleading picture about the influence of government size on
the impact of investment on growth. It seems the inter-country differentials here
are more related to the country-specific details of government activity than to
the overall size of government.
Turning to the export growth impacts, again we get a picture that varies quite
a bit across countries in a manner that is not related to government size. Thus,
government size alone does not appear to matter much regarding the impact of
export growth on economic growth. Indeed, if one examines the export impacts,
one thing stands out — the largest (and statistically significant) impacts on
economic growth appear to be quite clearly associated with the small countries
(as measured by population). Thus, the impacts are largest for Honduras, Costa
Rica, Nicaragua and Guatemala. They are not only smallest for the largest
countries (e.g. the US and Mexico), but are also not significant for most of
them.

Summary and concluding remarks

In this article, we have attempted to investigate the role played by the size of
government in determining the impact of major determinants of economic
growth for a group of 17 Western hemisphere countries within a production
function framework, but in a much more general manner than previous studies.
To this end, we classified the countries into four groups of low, intermediate,
intermediate-to-high, and high ranges in terms of the overall size of the
government sector. We also developed the estimating model so that it more
accurately represented the law relating the rate of economic growth to the rate
of investment and the rates of export and labour force growth. Indeed, in light
of the great diversity among countries, we used a random coefficients model,
which can be viewed as a generalisation of the laws stated by Pratt and
Schlaifer. Our results supported the random coefficients approach for the entire
sample. The results generally indicate that the role of capital formation and
export expansion in fostering economic growth is not adversely affected for
countries with large government sectors. Our estimates of the country-specific
Dar and Amirkhalkhali, Impact of Government Size on Economic Growth 75

estimates of the random coefficients, though very different across countries,

clearly support this finding for most countries considered in this study.

Appendix: Swamy’s random coefficients approach

Using matrix notations, Swamy's random coefficients estimator (1970) is:

B = (U0-1)-1U0-1Y

where B is the coefficient vector, Z is the (NTxK) matrix of observations on K

explanatory variables for N countries over T time periods, Y is an (NTx1)
vector on economic growth rates and 0 is the block diagonal covariance matrix
of disturbances. 0 is estimated on the basis of individual country time-series
regressions. The I-th diagonal block of estimated 0 is given by:

0ii = i iU+)ii
Note that Zi contains observations on K explanatory variables over T periods of
time for the ith country,

)ii = eiUei/(T-K), and = [((bibiU - (bi(biU/N)/(N-1)] - [()ii(ZiUZi)-1/N]

where ei = yi - Zibi, bi= (ZiU'Zi)-1ZiU yi and yi contains observations on the

economic growth rates over T periods of time for the ith country. We also use
the Swamy-Mehta (1975) estimator to obtain country-specific estimates of the
random coefficients. The country-specific random coefficients, Bi are obtained
as:

Bi = B + ZiU(Zi ZiU + )iiI)-1(yi - ZiB)

We then test the null hypothesis Ho: B1 = B2 = ... = BN using Swamy's g-

statistic which is given by:

g = ([(bi - D)UZiUZi(bi - D)/)ii] ~ 32K(N-1) where D = (()ii-1ZiUZi)-1 ()ii-1ZiUZibi

References

Alam, M.S. (1991) ‘Trade Orientation and Macroeconomic Performance in

LDCs: An Empirical Study’, Economic Development and Cultural Change
39: 839–48.
76 Development Policy Review

Alesina, A. and Perotti, R. (1994) ‘The Political Economy of Growth: A Critical

Survey of the Recent Literature’, World Bank Economic Review 8 (3):
351–71.
Amirkhalkhali, S. and Dar, A. (1995) ‘A Varying-Coefficients Model of Export
Expansion, Factor Accumulation, and Economic Growth’, Economic
Modelling 12 (4): 435–41.
IMF (1988, 1998) International Financial Statistics Yearbook. Washington,
DC: IMF.
IMF (1993, 1997) Government Finance Statistics Yearbook. Washington, DC:
IMF.
Jones, C.I. (1995) ‘Time Series Tests of Endogenous Growth Models’,
Quarterly Journal of Economics 110: 495–525.
Landau, D. (1983) ‘Government Expenditure and Economic Growth: A Cross-
Country Study’, Southern Economic Journal 49: 783–92.
Mankiw, N.G., Romer, D., and Weil, D. N. (1992) ‘A Contribution to the
Empirics of Economic Growth’, Quarterly Journal of Economics 107:
407–37.
Pratt, J. W. and Schlaifer,R. (1988) ‘On the Interpretation and Observation of
Laws’, Journal of Econometrics 39: 23–52.
Ram, R. (1986) ‘Government Size and Economic Growth: A New Framework
and Some Evidence from Cross-Section and Time-Series Data’, American
Economic Review 76:191–203.
Ram, R. (1985) ‘Exports and Economic Growth: Some Additional Evidence’,
Economic Development and Cultural Change 33: 415–25.
Rubinson, R. (1977) ‘Dependency, Government Revenue and Economic
Growth, 1955–70’, Studies in Comparative International Development 12:
3–28.
Solow, R.M. (1956) ‘A Contribution to the Theory of Economic Growth’,
Quarterly Journal of Economics 70: 65–94.
Swamy, P.A.V.B. (l970) ‘Efficient Inference in a Random Coefficients
Regression Model’, Econometrica 38: 3ll–23.
Swamy, P.A.V.B. and Mehta, J.S. (1975) ‘Bayesian and Non-Bayesian
Analysis of Switching Regressions and of Random Coefficient Regression
Models’, Journal of the American Statistical Association 70: 593–602.
Tanzi, V. and Zee, H.H. (1997) ‘Fiscal Policy and Long-Run Growth’, IMF
Staff Papers 44: 179–209.