Dynamic effects of energy sector public investment on sectoral economic growth: experience from Pakistan economy.
Ammad, Syed ; Ahmed, Qazi Masood
ABSTRACT
This study is an addition to a few studies in the developing
countries generally, and in Pakistan particularly, which aim at
investigating the effects of public investment in energy sector on
sectoral output, investment and employment. This study estimates the
dynamic elasticities of private investment with respect to public
investment to find crowding-out or crowding-in phenomenon in Pakistan,
and also to find out the long term marginal productivity and the share
of benefits. In addition the Study also reveals the changes in labour
absorption due to additional capital. The study covers eight sectors of
the Pakistan economy and uses the annual time series data from 1981 to
2011. Vector Auto Regressive (VAR/VECM) technique as developed by
Pereira (2000, 2001), which allows measuring the dynamic feedback effect
among the variables is used in the current study.
Twenty four sectoral elasticity coefficients for public investment
in energy sector are estimated. Of these seven out of eight confirm
crowding-in phenomenon in Pakistan economy. This overwhelming evidence
confirms that this public investment has positive effect on private
investment. The three out of eight elasticity coefficients show public
investment has increased labour absorption and remaining five show
labour is substituted by capital as a result of increased public
investment. Seven out of eight elasticity coefficients show positive
output effect, however the overall marginal productivities are lower
compared to several developing countries like Portugal and Spain where
such analysis has been conducted.
JEL Classification: C32, E62, H54, E22
Keywords: Public Investment, Economic Performance, Sectoral
Analysis, Pakistan
1. INTRODUCTION
The successive economic and financial crisis in recent time has
reemphasised the importance of fiscal policy. Modern literature has also
revisited the debate regarding the effectiveness of fiscal policy in
influencing growth. The issue of the impact of public investment on
growth is debated in economic literature since seminal work of Solow
(1955). The issue is tackled from different angles. Some have used
production function approach [Ligthart (2002), Otto and Voss (1994,
1996), Sturm and de Haan (1995) and Wang (2004)]. Then another seminal
work by Aschauer (1989) led a series of work on this issue once again in
empirical literature (1989a, 1989b). These approaches used single
equation method for estimation and captured only the direct effects of
public investment on growth. Periera (2000) gave another twist to this
literature by highlighting the indirect effects of public investment on
output through its effects on other inputs like private investment and
employment. Periera's works (1999, 2000, 2001, 2003, 2005, 2007 and
2011) also contributed empirically to this literature by using vector
autoregressive (VAR) technique. This work accounts for both the direct
and indirect effects of public investment on growth and also considers
the feedback effects of each input to other and finally their effects on
output.
The classical school believes that an increment in public spending
slows down growth and crowd out the private investment. Since higher
spending requires higher taxes at individual or corporate level, it
creates distortion in the choice of economic agents and increases
interest rate. Barro (1991) in his most famous work associated with
government size found a negative relationship between growth and
government size. Razzolini and Shughart (1997) in the case of United
States found a negative relationship between growth rate and relative
size of government. Parker (1995) in case of India found crowding out
effect of overall public investment while infrastructure investment
crowd in private investment. Alesina, et al. (2002) measured the effect
of fiscal spending in case of OECD countries in a Tobin's Q model
and confirmed a crowding out phenomena. Many other empirical studies
found evidence of crowding out effect of government expenditures
including [Ganelli (2003), Voss (2002), Engen and Skinner (1992),
Folster and Henrekson (2001), Devarajan, et al. (1996), Milesi and
Roubini, (1998) and Majumdar (2007)].
The Keynesians on the other hand, consider government spending as a
key variable for economic growth. They argue that development
expenditures on health, education and infrastructure increase labour
productivity and reduce cost of business, which motivates private
investment. Many empirical studies support this view. For instance like
Chakraborty (2007) examined the real and financial crowding out effect
in India using data from 1971 to 2003 through a VAR model and found that
public and private investment are complementary. Easterly and Rebelo
(1993) in their work found a positive growth effect of public
investment, specially transport and communication. Baotai (2004)
analysed the effect of public investment through cointegration model
during the period 1961 to 2000 for Canada and found mixed results; some
public expenditure such as health and education have a positive effect
while infrastructure and social security have a negative growth effect.
Bose, Haque and Osborn (2007) using data for 30 developing countries
found out that government capital expenditures have a positive effect on
growth, while at the disaggregate level only education expenditures are
positively correlated with growth.
Pereira (2000) investigated the effects of aggregate public
investment and infrastructure investment at a disaggregate level by
using the VAR model for U.S and found that both at aggregate and
disaggregate levels, public investment positively affects output and
crowd in private investment. This study estimated a marginal
productivity of 4.46 indicating that a one dollar investment will
increase private output by about $4.46 and found out that the highest
rate of return is in electric, gas, transit system and airfield sectors.
Pereira and Oriol (2001) analysed the marginal productivity of
private investment, output and employment with respect to public
infrastructure investment in the case of Spain by using VAR methodology.
The study used five VAR models, one for aggregate level and remaining
four for agriculture, services, manufacturing and construction. The
results indicate that at aggregate level public infrastructure
investment has positive marginal productivity for each variable while at
sectoral level manufacturing, services and construction have positive
output, private investment and employment marginal productivity but in
the case of agriculture there is negative marginal productivity of
output, private investment and employment. The highest output marginal
productivity was found in the case of manufacturing being 2.43
indicating one peseta of public investment will generate 2.43 pesetas of
output.
Pereira and Andraz (2005) analysed the effect of aggregate public
transportation infrastructure investment and its components (national
roads, municipal roads, highways, ports, airports and railways) on
aggregate private investment, aggregate output and employment in
Portugal by using a VAR approach on annual data from 1976 to 1998. They
found out that in the long term, aggregate public infrastructure
investment of one euro will generate an output of 9.5 euros and also
have a positive effect on private investment and employment. At a
disaggregate level, they found similar trends for output, employment and
revenue. Pereira and Sagales (1999) using the VAR model for Spain found
a crowding in effect of public capital on private output and employment.
Pina and Aubyn (2006) examined the rate of return of public investment
in the case of U.S economy using VAR model for a period of 1956-2001.
The four variables used were real private investment, real public
investment, private employment and real GDP and found a positive
Partial-cost dynamic feedback rate of return of 7.33 percent while the
total or Full-cost dynamic feedback came out to be 3.68 percent.
Pereira and Pinho (2011) using the data of twelve euro-zone
countries for 1980 to 2003 employed the same methodology and found
diverse results. For example, they established that public investment
has a positive effect on private investment and employment in all
countries except Austria, Belgium Luxembourg and Netherland, while
public investment has a positive effect on output in all countries
except Luxembourg and Netherland. They also concluded that in the case
of Austria, Belgium, Luxembourg and Netherland the public investment has
a negative output affect. But in Finland, Portugal and Spain public
investment has a positive growth effect; still it is unable to generate
sufficient tax revenue. While in case of France, Greece and Ireland
public investment pays for itself and finally in the case of Germany and
Italy, public investment not only pays for itself but also generates
extra tax revenue.
Afonso and Aubyn (2008) utilised accumulated impulse response
function of VAR model, which consists of real interest rate, real
output, real taxes, real public investment and real private investment
for 14 European Union countries and some non-European countries
including Japan, Canada and the United States. The results show that
output elasticity of private investment is higher than public
investment. Further in most of the countries they found a positive
marginal productivity accompanied with a crowd-in effect. Voss (2002)
investigated the crowding in or out effects in case of Canada and U.S
using quarterly data through a VAR model, using real GDP, real interest
rate, and share of public and private investment in the GDP. In both
countries he found a negative effect of public investment on private
investment. Mittnik and Neumann (2001) examined the relationship between
public investment, private investment and output using the VAR model for
six industrial countries. Results reveal that public investment crowd in
private investment in three countries only; however the public
investment has a positive output effect in all six countries.
Kamps (2005) measured the elasticites of private investment,
employment and output with respect to public investment using a VAR
estimation technique based on the variables: "net public capital
stock", "number of employed persons", "real
GDP" and "private net capital stock". The study was based
on 22 countries and showed that public capital stock has a positive
effect on output in majority of the countries excluding Japan and
Portugal. Further public investment and private investment are
complementary and crowding in exists except for Belgium, Japan and U.S.
However in the case of employment there is no significant role of public
capital.
Pereira (2001) estimated the VAR model using private gross domestic
product; private investment, public investment and private employment
for U.S economy and both private and public investment are further
disaggregated into highways and streets, electric and gas facilities,
sewage, water supply, education, hospital building and development
structure. At aggregate level he found that public investment has a
positive effect on private investment, the marginal productivity was
$4.5 with an annual rate of return of 7.8 percent. Pereira and Andraz
(2003) examined the effect of aggregate public investment on aggregate
private output, employment and investment in the case of U.S using VAR
impulse response methodology and found at aggregate level, public
investment exerts positive effect on all variables. The study found that
an investment of one million dollars will generate 27 new jobs in the
long term and one dollar investment of public investment will create
$1.112 of private investment and $4,991 of output with an annual rate of
return of 8.4 percent. Pereira and Andraz (2003) further analysed the
effect of aggregate public investment at disaggregate level and found in
six out of twelve industries public investment has a positive employment
effect; in five industries crowding in prevailed, while in eight out of
twelve industries, public investment has a positive effect on output.
Hyder (2001) examined the effect of real public investment on
private investment and growth through a VEC model during 1964 to 2001
and found a complementary relationship between public and private
investment and positive growth effect. Saeed et. al (2006) examined the
effect of public investment at aggregate and disaggregate level in a VAR
model using the variables i.e. public investment, employed labour force,
GDP and private investment. The study reveals that in agriculture there
is crowding in effect while in manufacturing there is crowding out
effect and at the aggregate level the evidence is inconclusive. For
example Hussain, et al. (2009) found that defense and debt servicing
crowd out investment while development expenditures crowd in investment.
Naveed (2002) showed that public capital formation has a crowding in
effect. Haque and Montiel (1993) found a crowding out effect in case of
Pakistan.
The impact of aggregate public investment on growth is examined
vastly in the economic literature. This paper captures both the direct
and indirect effects of public investment in energy sector on sectoral
output, private investment and employment. This will highlight first the
size of the impact of public energy investment on sectoral output and
second its impact on private investment. This study also indicates which
sector of Pakistan's economy is getting most benefit of energy
investment. This will be useful information for the policy-makers.
The remaining study is organised as follows: Section 2 illustrates
methodological framework, Section 3 gives data and diagnostic test,
Section 4 is based on empirical results and finally conclusions and
policy implications are presented in Section 5.
2. METHODOLOGICAL FRAMEWORK
The selection of the methodology and the variables for the present
study are based on the empirical studies such as Pereira (2000) and
Kamps (2005); where a Vector Auto Regressive (VAR/VECM) technique is
used for measuring the dynamic effects of public investment. This
methodology significantly differs from the one used in the previous
studies related to Pakistan, although some studies applied Vector Auto
Regressive (VAR/VECM) models, yet their findings are based on error
correction term; other studies measured causality among public
investment, private investment and output or their results are merely
based on impulse response graphs for measuring the nature of effects
either positive or negative. For our analysis, we have divided
Pakistan's economy into the following sub sectors; Agriculture,
Manufacturing (large and small scale), Mining and Quarrying,
Construction, Electricity and Gas Distribution, Transport Storage and
Communication, Finance and Insurance plus Ownership of Dwellings and
Public Administration, Defence and Community Services. Hence, total
eight VAR models are estimated; one for each of eights sectors. The VAR
model corresponding to each sector is specified as follow:
[X.sub.t] = C + [p.summation over (i=1)] [A.sub.t] [X.sub.t-i] +
[[epsilon].sub.t] ... (2.1)
Where X is the vector of (4x1), C is the intercept vector also
(4x1), A is the matrix of coefficient (4x4) and [epsilon] is the vector
of error term. Each VAR model consists of Public sector energy
investment, Private investment, Output and employment for each sector.
The linear form of the model is
Xt = [DELTA]log lpub, [DELTA]log lpriv, [DELTA]log Y, [DELTA]log
Emp ... (2.2)
Where lpub, lpriv, Emp and Y are log of real public investment, log
of real private investment, log of real output and employment
respectively.
Dynamic Feedback Effects
For measuring the effect of public investment on other variables,
an impulse response function for each VAR model was generated. By
definition an impulse response function measures the effect of a shock
in an endogenous variable due to other variables in the model. It is
known that residual of the VAR are contemporaneously correlated. For
measuring the effect of shock in one variable due to other variable,
these residuals should be uncorrelated. The VAR model is modified in
such a way that contemporaneous correlation among the residuals is
diagonal, called orthogonalisation. To attain these uncorrelated
residuals, Choleski decomposition is used and accumulated impulse
response is calculated to measure the cumulative response of all
variables due to innovation in policy variables i.e. Public investment
in energy. The outcome of accumulated impulse response function provides
the accumulated long term elasticity of the selected variables due to
shock in policy variable where the long term is defined as the time
period in which shock disappeared.
Long Term Accumulated Marginal Productivity
The long term accumulated marginal productivity of policy variable
measures the unit change of the dependent variable due to one unit
change in policy variable. This concept of marginal productivity is
different from the conventional concept. One of the main distinctions is
that it is not based on the assumption of ceteris paribus; it refers to
the accumulated marginal product and captures all the dynamic feedback
among the variables. The value of marginal productivity is obtained by
multiplying the accumulated long term elasticity with the ratio of
policy variable to the response variable.
[[epsilon].sub.IPUB] = [DELTA] log [Y.sub.i]/[DELTA] log
[IPub.sub.i] ... (2.3)
The above Equation (2.3) is the long term elasticity, which is
obtained directly from an accumulated impulse response function against
each sector; which measures the accumulated change in growth rate of
different variables. The numerator is the accumulated change in output
growth rate of the ith sector, while the denominator is the accumulated
change in growth rate of public investment in the ith sector.
The above elasticity is transformed into long term marginal
productivity by using following formula
MP [equivalent to] [DELTA]Y/[DELTA]IPub = [[epsilon].sub.IPUB]
[Y.sub.i]/[IPub.sub.i] ... (2.4)
In this fashion for each sector; marginal productivities of private
investment, output and employment (in terms of number of jobs creation)
are measured.
3. DATA SOURCES AND DESCRIPTION
This study is based on annual time series data from 1981 to 2011
obtained from the State Bank of Pakistan Annual Report, 50 Years of
Pakistan Economy and various issues of Economic Survey of Pakistan. All
variables are converted into real terms based on 1999-2000 prices (2)
and their first differences in log form are used in the analysis.
Univariate Analysis
Stationarity of each variable is one of the necessary conditions
for forecasting using the VAR model and if there is cointegration then
the order of integration must be the same. Augmented Dickey-Fuller
(1979) and Philips Perron (1988) test are used to check the order of
integration. The final decision based on Philips Perron test results
reported in Table 1 show (3) that all the variables are non-stationary
at levels using a 5 percent confidence interval, except three variables,
which are level stationary. However, at first differences, all the
variables are stationary.
VAR Order Selection
Appropriate number of lags is a crucial decision for VAR
estimation. There are different information criteria available for
choosing a more parsimonious model and we have applied Schwarz (1978)
information criterion (SC) and Akaike (1974) information criterion
(AIC). For each model lag selection was made on the basis of Schwarz
information criterion. The results reveal (4) that in most cases one lag
is showing minimum information criterion value while maximum of four
lags were incorporated to avoid too many parameters.
Diagnostic Test
The results of the diagnostic tests are given in Table 2. The
results indicate that there is no Heteroskedasticity in any model. The
results of LM test also support no serial correlation in all the cases
except services sector model. The assumption of Normality is also tested
in all the cases and the results do not support the normality
assumptions in five out of eight cases, but we can ignore this issue as
Lutkepohl (1991) discussed that the VAR parameters estimators do not
depend on the normality assumption.
Cointegration Analysis
Finally, to decide whether to use Vector Autoregressive Model (VAR)
or Vector Error Correction (VEC), a cointegration test is applied to all
the models by using Engle-Granger (1987) and Johansen (1991, 1995)
approaches. The cointegration results based on Engle-Granger test (5),
in all the models reject the existence of cointegration, while in a few
models only Johansen test shows the existence of cointegration. The
reason for using Engle-Granger approach is based on the finding of
Gonzalo and Lee (1998) and Gonzalo and Pitarakis (1999) who mentioned
that Johansen approach has small sample bias for cointegration when it
does not exist. These findings are similar to other related studies e.g.
in the case of Portugal, Pereria and Andraz (2005) and in the case of
U.S, Pereria and Andraz (2003) did not find any cointegration.
4. EMPIRICAL RESULTS
This section discusses the empirical effects of public energy
investment on sectoral output, private investment and employment. These
effects are based on accumulated impulse response function. The effect
of a shock in public energy investment on sectoral GDP is traced in
terms of output elasticities. The effect of a shock in public energy
investment on sectoral private employment is traced in terms of private
investment elasticities, similarly the effects of a shock in public
energy investment on employment are measured in terms of employment
elasticities.
Table 3 gives summary of results of the impact of public investment
on output, private investment and employment and detailed graphs are
given in Appendix-A which are based on accumulated impulse response
function with a time horizon of 20 years. These unit shock effects of
public energy investment on output show that public energy investment
has a positive effect on the output of all sectors except electricity
and gas distribution sector. In case of private investment the impulse
response functions indicate that public energy investment also has a
positive effect on private investment in all the sectors except finance
and insurance, while in case of employment the impulse response function
graphs show that only three sectors out of eight have a positive
employment effect with respect to public energy investment. One more
important feature of these graphs, which is worth mentioning here is
that in all the cases the shocks effect dies out after five years,
except three sectors.
Measuring the Long-term Accumulated Effect of Public Capital
Formation
The Effects of Public Investment on Output
The effect of public investment on sectoral output is presented in
Table 4. The results indicate that public investment has positive output
effects for all the sectors except electricity and gas distribution. The
result shows the sum of marginal productivities across the sectors is
3.57 i.e., one rupee public investment will collectively generate the
output of rupees 3.57, which is low as compared to the relatively
advanced countries, such as in Spain; Pereira and Oriol (2001) found the
aggregate marginal productivity for output of 5.5, similarly in the case
of Portugal; Pereia and Andraz (2007) found aggregate marginal
productivity of output of 8. On the sectoral level, the public
investment's highest benefit share goes to manufacturing followed
by mining and quarrying, transport and communication, services,
agriculture, finance and insurance and then construction. The share
distribution is 24 percent, 21 percent, 17 percent, 11 percent, 10
percent and 3 percent respectively.
The Effects of Public Investment on Private Investment
Table 4 also discusses the impact of public investment on private
investment. The empirical results show that public investment has a
positive impact on private investment supporting the hypothesis of
crowding-in; in seven out of eight sectors i.e. except the services
sector. The results show the sum of marginal productivities of private
investment across the sectors is 1.35 indicating one rupee public
investments will increase private investment by Rs 1.35. These results
show that overall impact of public investment on private investment is
also low in Pakistan as compared to the other countries. In the case of
Spain Pereira and Oriol (2001) found the aggregate marginal productivity
of private investment is 10.18, similarly in the case of Portugal,
Pereia and Andraz (2007) found aggregate marginal productivity is 9.45.
On the sectoral level, the highest benefit share of public energy
investment goes to manufacturing followed by agriculture, services,
transport and communication, mining and quarrying, electricity and gas
and then construction. The share distribution is 47 percent, 11.5
percent, 11 percent, 6 percent, 6 percent and 5 percent respectively.
The Effects of Public Investment on Employment
The employment effect of public investment is presented in Table 4.
On the sectoral level, public investment has positive employment effect
in agriculture, construction and electricity and gas. The one million
rupees public investment will create highest employment in agriculture
sector followed by construction and then electricity and gas. In
comparison with other studies such as in the case of Portugal, Pereia
and Andraz (2007) found the highest benefit share of infrastructure
investment in the case of construction followed by finance, services,
and real estate. These results show in many sectors it is negative,
however these results are also consistent with other studies. For
example Pereira and Andraz (2007) found negative employment effect of
public infrastructure investment in agriculture, food, textile, other
manufacturing and real estate sectors in the case of Portugal.
5. CONCLUSION AND POLICY IMPLICATION
The objective of this study is to find empirical evidence of the
effectiveness of public energy investment in Pakistan. In literature,
usually the production function approach is applied for such analysis
while this study uses the VAR methodology which allows capturing dynamic
feedback effect of public investment on private investment, employment
and output.
The study is one of the pioneer attempts on the subject by
estimating the long term marginal productivities of public investment at
sectoral level. The study uses data of eight sectors of Pakistan economy
from 1981-2011. The study estimates eight elasticity coefficients to
investigate the impact of public investment on sectoral private
investment and confirms crowding-in phenomenon in seven out of eight
sectros in Pakistan's economy. This overwhelming evidence confirms
that public investment has positive a effect on private investment. The
three out of eight elasticity coefficients show public investment has
increased labour absorption and the remaining five show labour is
substituted by capital as a result of increased public investment. The
highest marginal productivity is 0.88 in manufacturing followed by 0.766
and 0.61 in mining and quarrying and transport and communication
sectors. This implies one rupee public investment in these sectors will
generate rupees 0.88, 0.766 and 0.61 in these sectors respectively.
Generally the marginal productivity is lower as compared to several
developed countries like Portugal and Spain where such analysis has been
conducted.
The results of this study provide the answers to some important
policy questions and also help in formulating future policy. This study
calculates the marginal productivities, which are useful in project
evaluation and investment decisions. The positive output effect
indicates that public energy investment is growth stimulating through
its direct effect and indirect effects.
Syed Ammad <ammadsyed@yahoo.com> is PhD Research Fellow,
Department of Economics, University of Karachi. Qazi Masood Ahmed
<qmasood@iba.edu.pk> is Professor/Director, Centre for Business
and Economics Research, Institute of Business Administration, Karachi.
APPENDIX-A
Impulse Response Graphs
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
[FIGURE 3 OMITTED]
REFERENCES
Afonso Antonio and Aubyn Miguel St (2008) Macroeconomic Rates of
Return of Public and Private Investment Crowding-in and Crowding-Out
Effects, ECB. (Working Paper Series No 864).
Akaike H. (1974) A New Look at the Statistical Model
Identification. IEEE Transactions on Automatic Control, Ac-19, 716-723.
Alesina, A., S. Ardagna, R. Perotti, and F. Schiantarelli (2002)
Fiscal Policy, Profits, and Investment. The American Economic Review
92:3, 571-89.
Alfred, M. Pereira and Oriol Roca-Sagales (2001) Infrastructures
and Private Sector Performance in Spain .Journal of Policy Modeling 23:
4, 371-384,
Aschauer, D. (1989a) Is Public Expenditure Productive? Journal of
Monetary Economics 23:2, 177-200.
Aschauer, D. (1989b) Does Public Capital Crowd Out Private Capital?
Journal of Monetary Economics 24:2, 171-188.
Baotai, Wang (2004) Effectiveness of Government Expenditure on
Private Investment: Canadian Empirical Evidence. Empirical Economics 30:
2, 493-504.
Barro, R. (1991) Economic Growth in Cross-Section of Countries.
Quarterly Journal of Economics 106:2, 407-443.
Bose, Niloy, M. Emranul Haque and Denise R. Osborn (2007) Public
Expenditure and Economic Growth: A Disaggregated Analysis for Developing
Countries. The Manchester School 75: 5, September.
Chakraborty, Lekha (2006) Fiscal Deficit, Capital Formation and
Crowding Out: Evidence From India. (NIPFM Working Paper 06/43).
Devarajan, S., V. Swaroop, and H. Zou (1996) The Composition of
Public Expenditure and Economic Growth. Journal of Monetary Economics
37, 313-344.
Dickey, D. A. and W. A. Fuller (1979) Distribution of the
Estimators for Autoregressive Time Series With A Unit Root. Journal of
the American Statistical Association 74, 427-431.
Easterly, W. and S. Rebelo (1993) Fiscal Policy and Economic
Growth: An Empirical Investigation. Journal of Monetary Economics 32,
417-458.
Engen, Em and J. Skinner (1992) Fiscal Policy and Economic Growth.
(NBER Working Paper 4223).
Engle, R. F. and C. W. J. Granger (1987) Cointegration and Error
Correction Representation: Estimation and Testing. Econometrica 55,
251-276.
Folster, S. and M. Henrekson (2001) Growth Effects of Government
Expenditure and Taxation in Rich Countries. European Journal of
Political Economy 45, 1501-1520.
Ganeli, Giovanni (2003) Useful Government Spending, Direct
Crowding-Out and Fiscal Policy Interdependence. Journal of International
Money and Finance 22, 87-103.
Gonzalo, J. and T. Lee (1998) Pitfalls in Testing for Long Run
Relationships. Journal of Econometrics 86, 129-54.
Gonzalo, J. and J.-Y. Pitarakis (1999) Dimensionality Effect in
Cointegration Analysis. In R. Engle and H. White (eds) Festschrift in
Honour of Clive Granger. Oxford: Oxford University Press
Haque, N. U. and P. J. Montial (1993) Fiscal Adjustment in
Pakistan: Some Simulation Results. International Monetary Fund Staff
Papers 40:2, 471-80.
Hussain Adnan, Muhammad Sulaiman D., Akram Kamran and Lai Irfan
(2009) Effectiveness of Government Expenditure Crowding-In Or
Crowding-Out: Empirical Evidence in Case of Pakistan. European Journal
of Economics, Finance and Administrative Sciences 16.
Hyder, Kalim (2001) Crowding-Out Hypothesis in a Vector Error
Correction Framework: A Case Study of Pakistan. The Pakistan Development
Review 40: 4, 633-650.
Johansen, S. (1991) Estimation and Hypothesis Testing of
Cointegrating Vectors in Gaussian Vector Autoregressive Models.
Econometrica 59, 1551-1580.
Johansen, S. (1995) Likelihood-Based Inference in Cointegrated
Vector Autoregressive Models. Oxford: Oxford University Press.
Kamps, Christophe (2005) The Dynamic Effects of Public Capital: VAR
Evidence For 22 OECD Countries. International Tax and Public Finance,
Springer 12:4, 533-558.
Ligthart, J. (2002) Public Capital and Output Growth in Portugal:
An Empirical Analysis. European Review of Economics and Finance 1: 2,
3-30.
Lutkepohl, H. (1991) Introduction to Multiple Time Series Analysis.
Berlin: Springer.
Majumder, Md. Alauddin (2007) Does Public Borrowing Crowd-Out
Private Investment? The Bangladesh Evidence. Policy Analysis Unit
Research Department Bangladesh Bank (Wp#0708).
Milesi-Ferretti, G. and N. Roubini (1998) On the Taxation of Human
and Physical Capital in Models of Endogenous Growth. Journal of Public
Economics 70, 237-254.
Mittnik, S. and T. Neumann (2001) Dynamic Effects of Public
Investment: Vector Autoregression Evidence from Six Industrialised
Countries. Empirical Economics 26, 429-446.
Naqvi, Naveed H. (2002) Crowding-in or Crowding-out? Modeling the
Relationship between Public and Private Fixed Capital Formation. The
Pakistan Development Review 41:3, 255-76.
Otto, G. and G. Voss (1994) Public Capital and Private Sector
Productivity. The Economic Record 70: 209, 121-132.
Otto, G. and G. Voss (1996) Public Capital and Private Production
in Australia. Southern Economic Journal 62: 3, 723-738.
Parker, K. (1995) The Behaviour of Private Investment. Washington,
D.C.: International Monetary Fund. (IMF Occasional Paper No. 134).
Pereira, Alfredo M. and Pinho Maria De Fatima (2013) On the
Economic Effects of Public Infrastructure Investment: A Survey of the
International Evidence. The College of William and Mary. (Working
Paper).
Pereira, Alfredo M. and Pinho Maria De Fatima (2011) Public
Investment, Economics Performance and Budgetary Consolidation: VAR
Evidence for the 12 Euro Countries. Journal of Economic Development
36:1, 1-20.
Pereira, Alfredo M. and Jorge M. Andraz (2003) On the Impact of
Public Investment on the Performance of U.S Industries. Public Finance
Review 31:1, 66-90.
Pereira, and Sagales (1999) Public Capital Formation and Regional
Development in Spain. Review of Development Economics 3:3, 281-294.
Pereira, A. (2000) Is All Public Capital Created Equal? Review of
Economics and Statistics 82:3, 513-518.
Pereira, A. and J. M. Andraz (2007) Public Investment in
Transportation Infrastructures and Industry Performance in Portugal.
Journal of Economic Development 32, 1-20.
Pereira, A. M. and J. M. Andraz (2005) Public Investment in
Transportation Infrastructure and Economic Performance in Portugal.
Review of Development Economics 9, 177-196.
Pereira, Alfredo M. (2001) Public Capital Formation and Private
Investment: What Crowds in What? Public Finance Review 29:1, 3-25.
Phillips, P.C.B. and P. Perron (1988) Testing for a Unit Root in
Time Series Regression. Biometrika 75, 335-346.
Pina, A. and M. St. Aubyn (2006) How should we Measure the Return
on Public Investment in a VAR? Economics Bulletin 8:5, 1-4.
Razzolini, Laura, and William F. Shughart Ii (1997) On the
(Relative) Unimportance of Balanced Budget. Public Choice 90, 215-233.
Saeed, Hyder and A. Ali (2006) The Impact of Public Investment on
Private Investment: A Disaggregated Analysis. The Pakistan Development
Review 45:4, 639-663.
Schwarz, G. (1978) Estimating the Dimension of a Model. The
Analysis of Statistics 6, 461-464.
Sturm, J. and J. De Haan (1995) Is Public Expenditure Really
Productive? New Evidence for the U.S.A. and the Netherlands. Economic
Modelling 12: 1, 60-72.
Voss, G. (2002) Public and Private Investment in the United States
and Canada. Economic Modeling 19, 641-664.
Wang, B. (2004) Effects of Government Expenditure on Private
Investment: Canadian Empirical Evidence. Empirical Economics 30,
493-504.
Comments
It is an awesome topic to work on in the current scenario because
the country is facing acute problem of energy which is among the major
input in industrial as well as agriculture production. While reading the
paper I felt that if authors can incorporate the following comments, it
would enhance the quality of their paper.
Authors have used Growth model. Mankiw, Romer Weil (1991) already
showed that human capital is extremely important in case of growth
modeling, therefore, human capital is extremely important to include in
the growth equation.
Since not all the sectors need energy such as finance and insurance
thus all the sectors do not need to regress on energy. Therefore, I
would recommend to exclude irrelevant variables from the analysis.
Moreover, investment in public sector energy ventures are the investment
in the manufacturing sector by the public sector, but rest of the
investment is missing in the model. The variable is extremely important
and should be included in the model to get correct partial association
with the main variables.
Paper did not explain procedure adopted to fill the gaps in
employment data. As a reader it is a useful information which is
missing.
Cointegration in case of growth equation may not be a feasible
technique because there are significant chances that labour, capital,
human capital and growth are interlinked to each other and there is a
problem of endogeneity. Therefore, proper technique should be applied to
get the parameters.
The exercise done in Tale 4 is a very good exercise. However, the
magnitude and signs of few variables seems to be incorrect. I believe
that by including the human capital variables, inclusion and exclusion
of relevant and irrelevant variables and adopting proper estimation
technique may help in getting correct signs.
As much as I am not convinced with the estimation technique applied
in the paper, I am also not convinced with the application of impulse
response function on annual data. Impulse response function gives us the
response of shock in any variable within the system. By using this
technique we know the divergent or converging behavior of the variables.
However, it also tells us the duration of period in which shock is
either absorbed or tells. Using the technique on annual data, mostly, do
not give meaningful results. Therefore, in my view either this technique
is not used on annual data or the results should be interpret with
caution because "variable will adjust after 8 periods implies 8
years", which in most of the cases is not a meaningful result.
M. Ali Kemal
Pakistan Institute of Development Economics, Islamabad.
(2) The data is available in real terms at different base years.
For this study as suggested by the discussant we have used a common base
of 1999-2000, for the conversion of the nominal variables into real
variables.
(3) Due to lack of space just Philips Perron results are reported,
but the complete results are available on demand.
(4) Due to lack of space results are not reported, but available on
demand.
(5) For the sake of brevity results are not reported, but available
on demand.
Table 1
Unit Root Test
Phillips-Perron Test Statistic
Level
With Trend and
Without Trend Intercept
Variable t-Statistic Prob. * t-Statistic Prob. *
LAgr_IPub -0.544194 0.8729 -1.961717 0.6065
LAgr_IPrv -0.771485 0.8178 -2.679558 0.2494
LAgr_Emp 1.355936 0.9986 -2.668833 0.2537
LMing_GDP -0.487884 0.8843 -2.191037 0.4833
LMing_IPrv 0.053368 0.9585 -1.956587 0.6092
LMing_Emp -2.396637 0.1481 -2.754807 0.2207
LMfg_GDP -0.292774 0.9181 -2.522159 0.3166
LMfg_IPrv -0.657962 0.8472 -1.986704 0.5933
LMfg_Emp -0.321594 0.9136 -1.962546 0.6061
LConst_GDP -2.153902 0.2254 -1.578453 0.7865
LConst_IPrv -1.263144 0.6389 -3.388271 0.0652
LConst_Emp -3.485632 0.0127 -5.753265 0.0001
LElec_GDP -3.033429 0.039 -1.417099 0.843
LElec_IPub -1.954775 0.3053 -1.363139 0.8589
LElec_IPrv -1.212813 0.6613 -1.613274 0.7726
LElec_Emp -2.104588 0.2439 -3.762389 0.0277
LTranp_GDP -0.911304 0.776 -3.171151 0.1027
LTranp_IPrv -0.737195 0.8271 -2.069132 0.549
LTranp_Emp -3.044822 0.038 -18.15966 0
LFinc_GDP -0.907251 0.7724 -2.47431 0.3375
LFinc_IPrv -1.352439 0.5923 -2.562142 0.2987
LFinc_Emp -1.937825 0.3114 -2.648321 0.2634
LSrv_GDP -1.509704 0.5201 -2.513062 0.3208
LSrv_IPrv -0.310469 0.9154 -2.38316 0.3832
LSrv_Emp -0.072283 0.9464 -6.040012 0
LAgg_GDP -1.01663 0.7399 -3.168162 0.1033
LAgg_IPrv -0.246937 0.9247 -2.376024 0.3868
LAgg_Emp 1.100535 0.997 -1.926615 0.6249
Phillips-Perron Test Statistic
First Difference
With Trend and
Without Trend Intercept
Variable t-Statistic Prob. * t-Statistic Prob. *
LAgr_IPub -9.31261 0 -9.993371 0
LAgr_IPrv -6.833569 0 -6.749098 0
LAgr_Emp -8.362981 0 -8.815865 0
LMing_GDP -6.817256 0 -6.751895 0
LMing_IPrv -7.043074 0 -7.235855 0
LMing_Emp -5.685598 0 -5.644688 0.0001
LMfg_GDP -5.750705 0 -5.68134 0.0001
LMfg_IPrv -5.112176 0.0001 -5.053197 0.0008
LMfg_Emp -6.843413 0 -6.833039 0
LConst_GDP -5.429063 0 -5.744962 0.0001
LConst_IPrv -10.32539 0 -10.17403 0
LConst_Emp -15.32939 0 -16.14105 0
LElec_GDP -7.213615 0 -9.89615 0
LElec_IPub -7.555604 0 -13.90007 0
LElec_IPrv -5.892388 0 -6.015573 0
LElec_Emp -12.33055 0 -12.90363 0
LTranp_GDP -6.598544 0 -6.506002 0
LTranp_IPrv -4.622056 0.0005 -4.566332 0.0034
LTranp_Emp -31.51532 0.0001 -33.01162 0
LFinc_GDP -5.001994 0.0003 -4.92316 0.0021
LFinc_IPrv -5.476395 0.0001 -5.471944 0.0005
LFinc_Emp -6.564159 0 -6.570572 0
LSrv_GDP -7.695887 0 -7.932222 0
LSrv_IPrv -6.381415 0 -6.31137 0
LSrv_Emp -16.19263 0 -15.71361 0
LAgg_GDP -10.29256 0 -9.94885 0
LAgg_IPrv -5.751953 0 -5.703555 0.0001
LAgg_Emp -6.48744 0 -6.597266 0
LAgr is representing the log of agriculture sector, Lming is
representing the log of mining sector, LMfg is representing the log
of manufacturing sector, Lconst is representing the log of
construction sector, Lelec is representing the log of electric and
gas sector, LTranp is representing the log of transport and
communication sector, LFinc is representing the log of finance and
insurance sector, LSrv is representing the log of services sector
and LAgg is representing the log of Aggregate economy.
EMP is representing the employment, IPub is representing the public
investment, Iprv is representing the private investment.
Table 2
Diagnostic Test: Dynamic impacts of Public Energy Spending
Numbers Autocorrelation
of Test
Sectors/Model Lags (p-value) (1)
Agriculture(Major and Minor Crops,
Livestock, Fishing and Forestry) 1 0.1958
Mining and Quarrying 2 0.5828
Manufacturing 1 0.3933
Construction 1 0.1936
Electricity and Gas Distribution 1 0.8288
Transport, Storage and Communication 1 0.5089
Finance and Insurance 1 0.5292
Services (Community Services, Public
Administration and Defense and
Ownership of Dwellings) 1 0.0019
Normality Heteroskedas-
Test ticity Test
Sectors/Model (p-value) (2) (p-value) (3)
Agriculture(Major and Minor Crops,
Livestock, Fishing and Forestry) 0.1381 0.6523
Mining and Quarrying 0.9435 0.5831
Manufacturing 0.145 0.9859
Construction 0.978 0.8569
Electricity and Gas Distribution 0 0.9359
Transport, Storage and Communication 0.766 0.8618
Finance and Insurance 0.001 0.5744
Services (Community Services, Public
Administration and Defense and
Ownership of Dwellings) 0.0017 0.1813
(1.) Based on VAR residual serial correlation LM test with null no
serial correlation.
(2.) Multivariate Jarque-Bera residual normality test. For the null
hypothesis of normality.
(3.) VAR Residual Heteroskedasticity Tests. For null hypothesis of
no Heteroskedasticity.
Table 3
Long Term Accumulated Impulse Response Effects
of Public Energy Investment
On On Private On
Sectors Output Investment Employment
Agriculture(Major Crops, Minor
Crops, Livestock, Fishing and
Forestry) + + +
Mining and Quarrying + + -
Manufacturing + + -
Construction + + +
Electricity and Gas Distribution - + +
Transport, Storage and Communication + + -
Finance and Insurance + - -
Services (Community Services, Public
Administration and Defense,
Ownership of Dwellings) + + -
Table 4
Effects of Public Energy Investment on Output, Private Investment
and Employment
Share Contribution
% of total %of total % of total
Output Private Employment
Sectors Investment
Agriculture 21.38 12.09 43.82
Mining and Quarrying 2.93 4.66 0.17
Manufacturing 18.09 25.55 13.42
Construction 2.35 1.45 6.24
Electricity and Gas
Distribution 2.33 2.62 0.7
Transport, Storage and
Communication 12.67 18.65 5.51
Finance and Insurance 4.48 4.70 0.91
Services 18.02 27.22 14.23
Sum 82.27 96.95 85
Elasticities
Output Private Employment
Investment
Sectors
Agriculture 0.0085 0.0640 0.0061
Mining and Quarrying 0.1220 0.0766 -0.1831
Manufacturing 0.0227 0.1025 -0.0190
Construction 0.0214 0.1884 0.0142
Electricity and Gas
Distribution -0.0074 0.1268 0.0038
Transport, Storage and
Communication 0.0227 0.0325 -0.0219
Finance and Insurance 0.0372 -0.1371 -0.0245
Services 0.0125 0.0237 -0.0356
Sum
Marginal Productivity
Output Private Employment
Investment
Sectors
Agriculture 0.3892 0.2107 3.0902
Mining and Quarrying 0.7666 0.0971 -0.3669
Manufacturing 0.8830 0.7132 -2.9306
Construction 0.1080 0.0746 1.0190
Electricity and Gas
Distribution -0.0370 0.0903 0.0302
Transport, Storage and
Communication 0.6172 0.1650 -1.3880
Finance and Insurance 0.3576 -0.1756 -0.2560
Services 0.4850 0.1754 -5.8241
Sum 3.57 1.35 -6.63
Shares of Benefits (%)
Output Private Employment
Investment
Sectors
Agriculture 10.79% 13.81% 74.65%
Mining and Quarrying 21.25% 6.36% --
Manufacturing 24.48% 46.73% --
Construction 3.00% 4.89% 24.62%
Electricity and Gas
Distribution -- 5.92% 0..73%
Transport, Storage and
Communication 17.11% 10 81% --
Finance and Insurance 9.91% -- --
Services 13.44% 11.49% --
Sum
Source: Authors' own estimation.
