Bismillaah
Abstract
The
objective of the study is to investigate the public investment effect on
private sector investment. Using the flexible accelerator model with annual
data for the period 1970-2006, this study considered the roles of certain
government policy instruments such as public investment, exchange rate and the
availability of credit to private sector. The econometric approach addressed
the problem of spurious regression by testing the variables for stationarity.
The Engle-Granger cointegration test was also performed to determine the
long-term relationship between variables. Lastly, this study presented the
Engle-Granger error correction model to capture the short-run dynamics of
private investment.
The
results obtained from the error correction model showed that current public
investment had a negative and significant effect on private investment. The coefficient suggested that increasing
public investment by 10 percent will decrease private investment by 2.5 percent
in the same period. However, when public investment was lagged one
period it showed a positive and significant impact on private investment. The coefficient indicated that increasing public
investment by 10 percent will increase private investment by 2.8 percent within
one year. The control
variables, expected output, credit to private sector and the real exchange
rate, showed a positive and significant effect on private investment.
Keywords: Private Investment, Public
Investment, Crowding-Out, Indonesia、Error
Correction Model (ECM), Engle-Granger cointegration
I. INTRODUCTION
1. Background
The production
function states that capital and labor are the two most important factors of
economic output. Increasing the amount of labor and capital and technology
improvement will increase economic growth. In developing countries, capital’s
share of output is relatively low. In developed countries capital’s share of
output is around one third but in developing countries it is lower than that. Therefore,
increasing capital’s share of output through investment in developing countries
can accelerate economic growth. In the long run
investment plays an important role in economic growth. However, in the short run,
investment is the most volatile component of GDP.
Capital investment decisions can be taken by both the private
and public sectors. Productive investment by the private sector can promote
economic growth. Private investment not only provides many kinds of products
and services for consumers but also creates many opportunities for entrepreneurs
to expand their business. Moreover, private investment is vital for reducing
poverty by creating jobs for many people. Finally, the government benefits from
private investment through tax payments. The government is supposed to use
money from tax revenues to finance strategic public expenditures such as
infrastructure, education, and health.
Private investment decisions are affected by many factors
including government policy on fiscal and monetary matters. The government
plays a central role in the economy and there is much debate about whether that
role should expand, contract, or stay the same. The main rationale for
government intervention in the economy is market failure which means that
market economy fails to allocate scarce resources to the most efficient
outcome. On the other hand, government failure can also happen if government
policy fails to correct inefficiency in the market.
Many people
believe that private investment is more productive and efficient than public
investment. Private investment decision-making is based on the profit motive.
Investment decisions are made after considering many factors such
opportunities, competition, costs, risk, and return. On the other hand, public
investment is motivated by social welfare objectives. Social benefits and cost
analysis dominate in public investment decisions. In addition, private
investment is more sustainable than government investment as a source of
economic dynamism. Governments have limited sources to finance their investments.
To achieve certain economic growth, capital formation at certain levels has to
be achieved. Private investment has to fill the gap between the desired and
actual investment already undertaken by government investment.
Investment
decisions taken by government generally are to provide public goods. The
private sector generally is not interested in providing public goods because
public goods are not commercially feasible. So, public goods provision is also
a rationale for government intervention. The government also provides public
goods indirectly through state-owned enterprises. If the government provides
appropriate public goods such as infrastructure in efficient way, the
investment will have a positive impact on the private sector. Otherwise, it
will crowd out private investment. Competition over economic resources between
private and public sectors to finance their investments tends to crowd out the
private sector.
Private investment is vital for promoting economic growth and
reducing poverty. So understanding the factors that determine private
investment becomes a top priority for policy makers in formulating policies to
stimulate private sector development so that high economic growth can be
achieved and poverty can be reduced in Indonesia.
2. Problem Statement
Private investment is very crucial for economic growth so
that knowing economic factors that determine private investment is very
important for policy makers to formulate policies to encourage private
investment and achieve high economic growth, and reduce poverty.
The expected results of this analysis are the answers to the following research question: Has one unit of additional investment
by government or public sector reduced or increased private investment in the
recent past?
3. Objectives
The objective of this paper is to
construct a private investment model for Indonesia, to look into the determinants
of private investment in Indonesia with a main focus on the effect of public
investment, the availability of credit, and the exchange rate, and to provide
policy recommendations based on the empirical findings in order to stimulate
private investment in Indonesia.
4. Scope and
Limitations
Private investment is a wide subject and this paper intends to investigate
the determinants of gross fixed private investment in Indonesia, focusing on
public investment and other factors such as the availability of credit, and the
exchange rate. The investigation period is from 1970 to 2006.
5. Organization
This paper is organized into six parts. Part 1 is the introduction which presents what is
going to be discussed in this paper. Part 2 provides some economic background on the Indonesian
economy. The focus of this chapter is to review the development of private
investment, public investment, and public policy on private investment. Part 3 is
the literature review. This chapter discusses theoretical models of private
investment determinants and reviews some previous studies on private investment
determinants in developing countries. Part 4 provides information on data
sources and methodology to test the private investment determinants model. Part
5 is the main part of the paper which presents the empirical results and analysis. Part 6
concludes the paper and provides some policy recommendations.
II. OVERVIEW OF INVESTMENT
1. Investment Trend
Figure 2.1 shows the development of real gross fixed capital
formation as total investment, with private investment and public investment
shown separately in billion rupiah (constant 2000 prices) for the period
1970-2006. The gap between private and public investment was not large until
the 1980s. However, after the government
introduced its deregulation policy in 1983 and economic reform in 1986, the
difference became larger. The private sector became more dominant as an engine
of economic growth. Both total investment and private investment reached a peak
in 1997 at Rp431.2 and Rp359.2 trillion, respectively. After the Asian crisis,
investment took some time to recover from pre-crisis level. In 2002 public
investment attained the pre-crisis level although in 2005 it declined again and
was restored to a level higher than pre-crisis level in 2006.
Figure 2.1 Real Gross Fixed Capital
Formation (GFCF), Real Private Investment (PI) and Real Public Investment (GI)
(billions Rp constant 2000 prices)
Source: Central Statistics Agency,
Ikhsan and Basri (1991), Everhart and Sumlinski (2001), and World Bank (2007).
Table 2.1 shows the average annual growth rate of real
private and public investment for some specific periods. For the period
1970-1979 average annual total real investment growth was 13%. The growth rate
of real public investment was higher than that of private investment.
Government investment was still dominant during the period because the government
invested a large part of its oil revenues and foreign loans in development
projects. However, since 1982 government retrenched the annual state budget and
prioritized expenditures such as by reducing subsidies and cutting development
project expenditures.
During the period 1980-1989 both economic and investment
growth were slower than the previous period due to the economic recession in
1982 and the post oil boom era. In this period private investment grew faster than
public investment. The higher growth of private investment was due to a decreasing
government role in the economy. After the oil boom era was over, the government
introduced several deregulation policies and economic reforms such as the cancellation
of some development projects, banking and monetary liberalization in June 1983
to give freedom to state-owned banks to determine deposit and lending rates;
national tax reform in 1984; and economic reforms in 1986.
For the 1990s excluding the Asian financial crisis years
1997-1999, private investment’s growth rate was still higher than that of
public investment. The high growth of private investment was the result of
deregulation and economic reform during the 1980s. However, if the crisis years
are included in the calculation, private investment’s growth was negative but
public investment was still positive. This suggests that public investment had
a counter-cyclical effect.
Table 2.1 Average Annual Growth Rate
|
||||
Period
|
PI
|
GI
|
GFCF
|
GDP
|
1970-1979
|
10.6
|
18.1
|
13.0
|
7.0
|
1980-1989
|
9.7
|
2.8
|
7.1
|
4.8
|
1990-1996
|
10.1
|
2.7
|
8.4
|
6.2
|
1990-1999
|
-1.1
|
4.3
|
0.7
|
3.4
|
2000-2006
|
5.6
|
5.6
|
5.6
|
4.1
|
1970-2006
|
7.9
|
7.8
|
7.9
|
5.5
|
Rapid
growth 1971-1981
|
10.8
|
16.9
|
12.7
|
7.3
|
Oil
boom (1973-1982)
|
9.3
|
16.7
|
11.7
|
6.2
|
Adjustment
to lower oil prices 1982-1986
|
7.2
|
-3.5
|
3.5
|
3.9
|
Liberalization
and recovery 1987-1996
|
10.9
|
6.1
|
9.8
|
6.4
|
Asian
financial crisis (1997-1999)
|
-25.8
|
9.5
|
-17.6
|
-4.3
|
Note:
The formula for average annual growth rate is 
where n is number of years in certain period, EV is value at the end of period and BV is value at the beginning of period. PI = Real Private Investment, GI = Real Public Investment, GFCF = Real Gross Fixed Capital
Formation, and GDP = Real Gross
Domestic Product.
where n is number of years in certain period, EV is value at the end of period and BV is value at the beginning of period. PI = Real Private Investment, GI = Real Public Investment, GFCF = Real Gross Fixed Capital
Formation, and GDP = Real Gross
Domestic Product.
Source: Central Statistic Agency, Ikhsan
and Basri (1991), Everhart and Sumlinski (2001), World Bank (2007), and IMF
International Financial Statistic.
Figure 2.2 shows total investment, private and public
investment as percentage of GDP. The ratio of private investment to GDP
fluctuated and reached a peak in 1997. During the Asian financial crisis,
private investment fell to 15.5% and 11.1% in 1998 and 1999 respectively. Until
2006 the ratio was still lower than that before the crisis. On the other hand,
public investment fluctuated less than private investment. Public investment
peaked at 8.1% in 1982 and stabilized at around 5-7% of GDP.
Figure 2.2 Total Investment (GFCF), Private
Investment (PI) and Public Investment (GI) as % of GDP
Note:
Investment and GDP in real terms.
Source: Central Statistic Agency, Ikhsan
and Basri (1991), Everhart and Sumlinski (2001), and World Bank (2007).
III. LITERATURE REVIEW
1. Investment Model
There are four broad types of investment models, as
identified by Sakr (1993) and Ghura and Goodwin (2000). They are the
accelerator model, the expected profit model, the neoclassical model, and Tobin’s
q model.
The accelerator model was originally proposed by Clark
(1917), and consists of two variants, the fixed and flexible accelerator models
(Sakr 1993). The fixed accelerator model states that capital stock has a
positive relationship with the level of output (aggregate demand) and
investment is defined as 
, where 
is the desired capital
stock at time t, 
is the actual
capital stock at time t-1, andδis the rate of depreciation of
the capital stock.
, where is the desired capital
stock at time t, is the actual
capital stock at time t-1, andδis the rate of depreciation of
the capital stock.
A slight variation, the flexible accelerator model, which was
proposed by Goodwin (1951) and Chenery (1952), uses a partial adjustment
mechanism (Valadkhani 2004). This model emphasizes that the desired capital
increase does not occur immediately, but depends on the speed of adjustment,
which is influenced by some observable variables such as government investment,
domestic credit, and the real exchange rate (Sakr 1993). The flexible
accelerator model is preferred for handling data limitation in developing
countries (Ramirez, 1994).[1]
In the expected profit model, investment is determined by the
profit motive as the incentive to invest (Sakr 1993). This model states that the higher the
expected profit level, the higher the investment level, because a firm has
greater internally generated funds. According to Chirinko (1993), variables
that reflect profits such as current profits, retained profits, output, price
or sales can influence the level of investment.
The neoclassical approach which is formulated by Jorgenson
(1963 and 1971), Hall-Jorgenson (1967) and Jorgensen-Siebert (1968) expresses
the aggregate investment as the change in capital stock (the adjustment between
current and the desired capital stock) and a function of the user cost of capital
and output variable (Serven and Solimano 1989). The neoclassical model stresses
on the importance of the user cost of capital variable as the implicit price
that a firm has to pay to use capital goods to produce goods and services (Saxton
2002). The user of cost of capital depends on the capital asset price, the
depreciation rate, the real interest rate, and tax (Saxton 2002).
The Tobin’s q investment model is associated with Tobin
(1969). Tobin’s q is the ratio between the market value of capital stock and
its replacement cost, and is the main factor that drives investment (Serven and
Solimano 1989). Under Tobin’s q model, investment takes place if the share
value increases higher than the replacement cost of physical capital (Sakr
1993). Applying Tobin’s q model to developing countries such as Indonesia is
difficult because stock market valuations of firms’ fixed assets are
non-existent, or even if they exist, they do not reflect the firms’ market
value and future profitability (Shafik 1990).
The above models
have been applied successfully in industrial countries. However, two problems
will arise if we use those frameworks in developing countries: (1) some
assumptions may not hold, such as maximization of rates of return by economic
agents, the existence of perfect market for goods and well-developed financial
markets, and little or no government interventions; and (2) a lack of data on
the capital stock, real wages, and real financing rates for debt and equity
(Greene and Villanueva 1991).
So, the application
of the above model in developing countries needs some adjustments because some
of the assumptions of the model are not applicable in developing countries.
Hence, investment determinants in developing countries usually include
particular factors such as public investment and the availability of financing
as explanatory variables (Sakr 1993).
The
relationship between public investment and private investment could be
complementary or competitive (Chhibber and Dailami 1990). If public investment
expenditures result in the provision of public goods/services that reduce the
cost of production of the private sector, they will have a positive effect on
private investment. Public investment expenditures also raise aggregate demand
and increase capacity utilization in the private sector. On the other hand,
public investment expenditures can crowd out private investment in the competition
for limited financial resources. In developing countries, the private sector
depends highly on debt capital, particularly bank loans, so financial crowding
out is more serious (Chhibber and Dailami 1990).
The empirical study of the relationship between public
investment and private investment is still not conclusive. It depends on
observation periods and countries. Sundararajan and Thakur (1980) use a
combination of neoclassical and accelerator theories to elicit different
results between India and Korea. In India, government investment has a positive
effect on private investment in the long run but in the short run it crowds out
private investment. On the other hand, in Korea, government investment has a
positive effect on private investment both in the long run and in the short
run. A study by Wai and Wong (1992) in Greece, Korea, Malaysia, Mexico, and
Thailand shows inconclusive result. Using a modified version of the flexible
accelerator theory, they find that government investment has a positive and
significant effect on private investment only in Greece, but insignificance
effect for other countries (Korea, Malaysia, Mexico, and Thailand).
Blejer and Khan (1984) use the flexible accelerator model to
determine the effect of government policy on private investment. The result
shows that increasing government investment tends to decrease private
investment in the short run but has no effect in the long run.
Chhibber and van Wijnbergen (1988) use an accelerator model
to study private investment determinants in Turkey for the period 1970-86. The
four key variables as private investment determinants are capacity utilization,
credit to the private sector, the effective cost of borrowing, and the
composition of public investment. The findings show that capacity utilization
and infrastructure have a positive effect but not significant. The real cost of
borrowing and non-infrastructure have a negative and significant influence on
private investment.
Cruz and Teixeira (1999) examine the impact of public
investment on private investment in Brazil for the period 1947-1990, using an
Engle-Granger error correction model. They find that in the short term the
public investment crowds out private investment but in the long term it crowds
in private investment.
Erden and Holcombe (2005) follow Blejer and Khan (1984) and
Ramirez (1994)’s approach by using a flexible accelerator model to examine the
effect of public investment on private investment in developing countries.
Using a panel of 19 developing countries for the period 1982-1997, they find
that public investment has a positive effect on private investment whereas for
12 industrial countries for the period 1982-1996, public investment has a
negative impact on private investment.
This study constructs an empirical framework based on the
flexible accelerator model to reflect the institutional and structural
characteristics of Indonesia. The choice of the flexible accelerator model is
based on the reasoning that there is no publication of capital stock series or
reliable estimate for the rate of depreciation in Indonesia. The available data
is gross investment which also includes depreciation. Furthermore, there are
many economic distortions in Indonesia, although the Indonesian government has
tried to reduce the distortions through certain deregulations, liberalization,
and reforms in many sectors.
Regarding possible determinants of private investment, we limit
ourselves to the variables which we consider being most relevant and whose data
availability is not a serious problem. For example, we do not choose the real
interest rate variable because 14 data points out of 37 sample observations are
negative, which indicates financial repression. According to McKinnon (1973),
if financial markets are in financially repressed conditions, the quantity of
credit is more binding than the cost of credit.
IV. METHODOLOGY AND
DATA
1. Empirical Model
To estimate the flexible accelerator model, we follow Blejer
and Khan (1984), Ramirez (1994) and Erden and Holcombe (2005) to capture the
characteristics of developing countries like Indonesia. In the flexible
accelerator model, it is assumed that the desired capital stock is proportional
to the level of expected output
 |
where 
is the desired capital
stock of the private sector in period t and 
is the expected level
of output in period t. The actual
private capital stock does not adjust completely to reach the desired level due
to technical constraints and the time needed for planning, deciding, building
and installing new capital. Actual private capital stock adjusts to the difference
between desired capital stock in period t
and actual private capital stock in the previous period, so that we can specify
a partial adjustment mechanism as follows
is the desired capital
stock of the private sector in period t and is the expected level
of output in period t. The actual
private capital stock does not adjust completely to reach the desired level due
to technical constraints and the time needed for planning, deciding, building
and installing new capital. Actual private capital stock adjusts to the difference
between desired capital stock in period t
and actual private capital stock in the previous period, so that we can specify
a partial adjustment mechanism as follows
where 
is actual private
capital stock in period t and β is the coefficient of adjustment. The coefficient
of adjustment in the equation captures the response of private investment to
the gap between desired and actual investment. The coefficient is between 0 and
1, or
. The β coefficient
equal to 1 means adjust instantaneously to its desired level and no adjustment
if the β coefficient equal to 0. The β coefficient of adjustment is also assumed
to vary systematically with the economic factors that influence the ability of
private investors to achieve the desired level of investment. We hypothesize
that the response of private investment depends on public investment, the
availability of credit to private sector, and real exchange rate.
is actual private
capital stock in period t and β is the coefficient of adjustment. The coefficient
of adjustment in the equation captures the response of private investment to
the gap between desired and actual investment. The coefficient is between 0 and
1, or. The β coefficient
equal to 1 means adjust instantaneously to its desired level and no adjustment
if the β coefficient equal to 0. The β coefficient of adjustment is also assumed
to vary systematically with the economic factors that influence the ability of
private investors to achieve the desired level of investment. We hypothesize
that the response of private investment depends on public investment, the
availability of credit to private sector, and real exchange rate.
Equation (4.2) can be rewritten as follows:
As private capital stock data in developing countries is not
available, we use gross private investment instead of private capital stock.
Theoretically, gross investment in the private sector is defined as being equal
to net investment in the private sector plus depreciation of the previous
capital stock. On the other hand, net investment in the private sector is
defined as the difference between the actual stock of capital in period t and the actual stock in the previous
period t-1.
where KPt –
KPt-1 is net private investment, PIt is gross private investment and δ is depreciation rate of private capital stock.
In the standard lag-operator notation, equation (4.4) can be
rewritten as:
 |
where L is the lag
operator, 
Equation (4.5) can be rewritten:
 |
Substituting equation (4.6) into equation (4.3), we get:
Equation (4.7) has an advantage that it does not require any
information of actual capital stock and net private investment and can be
estimated using available gross investment data.
We generalize this flexible accelerator model by adding some
other variables that might also affect investment, such as public investment,
credit to private sector, real exchange rate, and economic crisis. Adding these
variables to equation (4.7) and using equation (4.1) we get:
 |
where GIV is
government investment, CPS refers to
credit to private sector, REX denotes
the real exchange rate and DUM is a
dummy variable for economic crisis years that takes on a value of 1 for
1997-1999 and 0 for other years.
Equation (4.8) can be estimated after an unobservable
variable is somehow estimated.
To estimate expected output , a common method is to fit an autoregressive process and
produce predicted values. This study uses a second-order autoregressive process
of real GDP for the period 1960-2006 as follows:
is somehow estimated.
To estimate expected output , a common method is to fit an autoregressive process and
produce predicted values. This study uses a second-order autoregressive process
of real GDP for the period 1960-2006 as follows:
where 
is real GDP and 
is dummy variable that takes on a value of 1 for 1997-1999
(economic crisis years) and 0 for other years.
is real GDP and is dummy variable that takes on a value of 1 for 1997-1999
(economic crisis years) and 0 for other years.
2. Econometric
Framework
The long-term relationship between economic variables is
important because in the long-run all economic forces are in equilibrium
(Harris and Sollis 2003). It is also important to distinguish between
stationary and non-stationary variables. Stationary variables contain either no
trend or deterministic (fixed) trends, while non-stationary variables contain
stochastic (random) trends (Harris and Sollis 2003). Many economic time series variables
are non-stationary and performing regression on non-stationary variables results
in spurious regression, meaning that the regression result will have no
economic meaning. However, it is possible to regress a non-stationary time
series variable on others if these variables are cointegrated (Harris and Sollis
2003). This regression represents their long-run relationship (Harris and
Sollis 2003).
Non-stationary variables have a unit root and testing for the
presence of a unit root in variables is essential to determine the order of
integration of variables. If the data has unit roots and the variables are
integrated of the same order, the next step for testing cointegration is to estimate
the long-run relationships. Using the Engle-Granger approach for testing
cointegration, the residuals obtained from regression of the long-term
relationship have to be tested for the presence of unit roots. Cointegration
among variables will exist if the residuals are stationary or integrated of
order zero, I(0). If the variables
are cointegrated, an error correction model (ECM) must exist which contains
both information on the long-run and the short-term characteristics of the
model (Harris and Sollis 2003).
a. Testing for Unit
Roots
Testing for unit roots is necessary to prevent spurious
regression. According to Harris and Sollis (2003) “if a series must be
differenced d times before it becomes
stationary, then it contains d unit
roots.” Unit root testing includes testing for the order of integration of the
series. A notation 
means that a series 
is said to be integrated of order d (Harris and Sollis 2003).
means that a series is said to be integrated of order d (Harris and Sollis 2003).
Two commonly employed methods for unit root testing are
Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP). The three possible formulations of ADF unit
root test that include three different combinations of the deterministic part
are as follows (Pfaf 2006):
where 
is the variable under
consideration, T is time trend, and k is the number of lagged differences to
capture any autocorrelation. If the value of ADF statistic is smaller (more
negative) than its critical value, the null hypothesis of nonstationary is
rejected.
is the variable under
consideration, T is time trend, and k is the number of lagged differences to
capture any autocorrelation. If the value of ADF statistic is smaller (more
negative) than its critical value, the null hypothesis of nonstationary is
rejected.
On the other hand, according to Pfaff (2006) “Phillips-Perron
test is a nonparametric test statistic that allows for weak dependence and
heterogeneity of the error process.” There are also three possible
specifications of the Phillips-Perron test as follows (Widarjono 2005):
If the value of PP statistic is smaller (more negative) than
its critical value, the null hypothesis of nonstationary is rejected.
b. Cointegration
Engle and Granger (1987) introduce the concept of
cointegration as follows: “If 
is a vector of
economic variables, long-run equilibrium occurs when
is a vector of
economic variables, long-run equilibrium occurs when 
.
In most time periods, deviations from long-run equilibrium
occur, and therefore;
.”
Then, Engle and Granger (1987) present the formal definition
of cointegration as follows: “The components of the vector are said to be cointegrated
of order d, b, denoted by 
, if (1) all components of are integrated of order d or I(d); and (2) there exists a vector αsuch that 
.” The vector αis
called the cointegrating vector (Enders 1995).
are said to be cointegrated
of order d, b, denoted by , if (1) all components of are integrated of order d or I(d); and (2) there exists a vector αsuch that .” The vector αis
called the cointegrating vector (Enders 1995).
Cointegration concept has two essential characteristics.
First, cointegration requires the variables have the same order of integration,
and second, if there are N variables in the model, it is possible that the
model has N-1 cointegrating vectors (Enders (1995) and Harris and Sollis
(2003)). For example, if two series are both integrated of order one, I(1), there can be one cointegrating
vector.
c. Engle-Granger
Two-Step Procedure
Engle and Granger (1987) propose the two-step procedure to
estimate the cointegrating vector αand
to model the dynamic behavior of I(d)
variables (Pfaff 2006). The Engle-Granger approach is used for testing
cointegration in single equations (Harris and Sollis 2003). The first step of
the procedure is to estimate the long-run relationship with variables in I(1) as follows:
(4.16)
where 
is error term. If the
residuals are stationary, the variables are cointegrated (Pfaff 2006)
and the second step is to specify an error correction model (ECM).
is error term. If the
residuals are stationary, the variables are cointegrated (Pfaff 2006)
and the second step is to specify an error correction model (ECM).
3. Data
The data used in this study is annual data obtained from
different sources. The sample is for the period 1970-2006. The period is
selected because this is the period for which data is available. The unit of
account of the data is in billions rupiah (national currency), except the real
exchange rate is in units of national currency per US dollar.
Following Chhibber and van Wijnbergen (1988), Sakr (1993),
Ramirez (1994), and Erden and Holcombe (2005), all variables are in real terms[2] and
are expressed in natural logarithms. The reasons for expressing the variables
in natural logarithms are that the estimated coefficients have an elasticity
interpretation; a model with dependent variable in log form tends to follow the
classical linear model assumptions; and a model in log form reduces the problem
of outliers (Wooldridge 2006).
Data for nominal gross domestic product (GDP) was obtained
from the IMF International Financial Statistics line 99b. Data for the GDP
deflator was obtained from the IMF International Financial Statistics line
99bip. Real GDP was obtained by deflating nominal GDP with the GDP deflator. The
source of gross fixed capital formation at 2000 constant price data is the
Central Statistics Agency.
Data for public investment is from four sources.
1.
The data for the period 1970-1977 is from Ikhsan
and Basri (1991). Data from this publication is government investment in 1983
constant prices. We have to adjust them to 2000 constant prices using CPI.
2.
The data for the period 1978-1985 is from
Investment Analysis published by Central Bureau of Statistics (BPS 1988). Data
is in current prices and has to be adjusted to 2000 constant prices using CPI.
3.
For the period 1986-1999, the series are from
Everhart and Sumlinski (2001). Data is presented in terms of percentage of
current price GDP. To get current price of public investment, we multiply the
percentage by current price GDP. To get the real public investment, we adjust
them by deflating with CPI.
4.
Data for the last period 2000-2006 is compiled
from the World Bank report’s Indonesia Public Expenditures Review 2007.
Data for public investment 1978-2006 consist of central and
local governments and SOE investment.
To measure private investment, this study follows the
approach of Everhart and Sumlinski (2001). Gross fixed capital formation in
national accounts usually is not divided into private and public sector. In
this study, public investment includes the development (capital) expenditures
of the central government and local governments and the capital expenditures of
the state-owned enterprises. Private investment is obtained by subtracting
public investment from gross fixed capital formation.
To obtain real expected output we follow the approach of
Blejer and Khan (1984), Ramirez (1994) and Erden and Holcombe (2005). To proxy
unobservable expected output, the common practice is to fit an autoregressive
process. Predicted values from the process are the expected output. Due to data limitation (annual
data), the possible process is a second-order autoregressive model of the real
GDP for the period 1960-2006:
(4.21)
where dum is a
dummy variable to capture the economic crises years that takes on a value of 1
for 1997-1999 and 0 for other years.
Data for banking credit to private sector was obtained from
Bank Indonesia. This nominal banking credit to private sector was deflated by
CPI to obtain real banking credit to the private sector.
Following Ramirez (1994), Ribeiro and Teixeira (2001), and
Erden and Holcombe (2005), the real exchange rate is domestic currency (rupiah)
nominal exchange rate against the US dollar inflated by the US CPI and deflated
by the Indonesian price levels. The real exchange rate is computed based on the
bilateral exchange rate between the rupiah and US dollar. Theoretically, the
best indicator is the real effective exchange rate. However, due to data
limitation this study uses the bilateral exchange rate. A nominal exchange rate
was obtained from the IMF International Financial Statistics line ae (market rate, rupiah per US dollar,
end of period). Data for Indonesian consumer prices index (CPI) was obtained
from the IMF International Financial Statistics line 64 (2000 = 100, period
averages). Data for the Unites States CPI was obtained from the IMF
International Financial Statistics line 64 (2000 = 100, period averages).
The depreciation rate (δ)of the private capital stock is assumed
to be 5%. It is arbitrarily chosen following the studies of Blejer and Khan
(1984), Ikhsan and Basri (1991), Ramirez (1994), and Erden and Holcombe (2005).
V.
EMPIRICAL RESULTS AND ANALYSIS
There are four steps in applying the Engle-Granger procedures
(Enders 1995):
1.
Testing the variables for their order of
integration (unit root test)
2.
Estimating the long-run equilibrium relationship
3.
Estimating the error-correction model
4.
Testing the model adequacy.
1.
Unit Root Tests
Unit root testing is necessary to determine the order of
integration. According to the Engle-Granger definition, cointegration requires
that the variables be integrated of the same order.
Two common methods to determine the number of unit roots are the
Augmented Dickey-Fuller (ADF) test and the Phillips-Perron (PP) test.
a.
Augmented Dickey- Fuller Test
The ADF test includes some lags of dependent variables to
eliminate serial correlation from the error terms. Choosing the optimal lag
length for the ADF test for a small sample is rather difficult. Too many lags
will cause the losing of a degree of freedom and too few lags will cause the
test to be incorrect (Wooldridge 2006). For annual data, Wooldridge (2006)
suggest using one or two lags. Alternatively, choosing the optimal lag length
can be based on the information criterion. In this study, the optimal lag length
that minimizes the value of the information criteria is used. Diebold (2007)
recommends using Schwarz information criteria (SIC) to select a parsimonious
model.
Table 5.1 shows the result of the ADF test together. The
figures are the ADF statistics. If the ADF
statistic is smaller than the critical values at the 1%, 5%, and 10%
levels; the variables are stationary. The ADF statistics in Table 1 indicate the
null hypothesis of nonstationary fails to be rejected if the variables are in
levels, although public investment is found to be stationary when only a
constant is included in the test regression model. But when the series are in
first-differenced the null hypothesis is rejected for all variables, which
suggests that we can assume all variables are integrated of order one.
b.
Phillips-Perron Test
To confirm the stationarity test results from ADF, the
Phillips-Perron (PP) test is conducted. The figures are the PP statistics. If
the PP statistics are smaller than
the critical values at the 1%, 5%, and 10% levels; the variables are
stationary. The PP test results are presented in Table 5.2 and the test
generally confirms the ADF test results, although public investment is found to
be stationary when only a constant is included in the test regression model.
Based on the ADF and PP tests and to avoid the tendency to over-reject
(under-reject) the null hypothesis of nonstationary when it is true (false)
(Harris and Sollis 2003), it is safe to say that the variables are integrated
of order one, I(1).
Table 5.1 Augmented
Dickey-Fuller Test Results
Unit-root tests at levels
|
|||
Variables
|
Constant
|
Constant
and Trend
|
None
|
Real private inv.
|
-1.878
(0)
|
-2.065
(0)
|
2.647
(0)
|
Real expected output
|
-2.001
(0)
|
-1.454
(0)
|
7.0533
(0)
|
Real public inv.
|
-2.905***
(0)
|
-2.405
(0)
|
2.489
(0)
|
Real credit
|
-0.955
(0)
|
-1.314
(0)
|
1.700
(0)
|
Real exchange rate
|
-1.057
(0)
|
-2.457
(0)
|
0.490
(0)
|
Unit-root tests at first differences
|
|||
Variables
|
Constant
|
Constant
and Trend
|
None
|
ΔReal private inv.
|
-5.377***
(1)
|
-5.576***
(1)
|
-4.500***(0)
|
ΔRealexpected output
|
-5.008***
(0)
|
-5.487***
(0)
|
-2.614**
(0)
|
ΔReal public inv.
|
-6.130***
(0)
|
-6.825***
(0)
|
-5.216***
(0)
|
ΔReal credit
|
-4.670***
(0)
|
-4.624***
(0)
|
-4.424***
(0)
|
ΔReal exchange rate
|
-5.567***
(0)
|
-4.840***
(1)
|
-5.621***
(0)
|
Notes:
-
***, **, and * denote significance at the 1%, 5%
and 10% level respectively and the rejection of the null hypothesis of
nonstationarity.
-
Figures in parentheses are automatic selection
of lag length based on the Schwarz information criterion (SIC), computed by
EViews 5.0 with maximum lag = 9.
-
All variables are expressed in natural
logarithms.
Table 5.2
Phillips-Perron Test Results
Unit-root tests at levels
|
|||
Variables
|
Constant
|
Constant
and Trend
|
None
|
Real private inv.
|
-1.878 (0)
|
-2.065 (0)
|
2.647 (0)
|
Real expected output
|
-2.000 (0)
|
-1.454 (0)
|
7.053 (0)
|
Real public inv.
|
-2.905* (0)
|
-2.405 (0)
|
2.489 (0)
|
Real credit
|
-0.955 (0)
|
-1.314 (0)
|
1.700 (0)
|
Real exchange rate
|
-1.057 (0)
|
-2.457 (0)
|
0.490 (0)
|
Unit-root tests at first differences
|
|||
Variables
|
Constant
|
Constant
and Trend
|
None
|
ΔReal private inv.
|
-5.711***
(0)
|
-5.955***
(1)
|
-4.500***
(0)
|
ΔReal expected output
|
-5.010***
(0)
|
-5.487***
(0)
|
-2.614**
(0)
|
ΔReal public inv.
|
-6.130***
(0)
|
-6.825***
(0)
|
-5.216***
(0)
|
ΔReal credit
|
-4.670***
(0)
|
-4.624***
(0)
|
-4.424***
(0)
|
ΔReal exchange rate
|
-5.567***
(0)
|
-5.835***
(1)
|
-5.621***
(0)
|
Notes:
-
***, **, and * denote significance at the 1%, 5%
and 10% level respectively and the rejection of the null hypothesis of
non-stationarity.
-
Figures in parentheses are automatic lag length selection
based on the Schwarz information criterion (SIC) using autoregressive OLS
spectral estimation method, computed by EViews 5.0 with maximum lag = 9.
-
All variables are expressed in natural
logarithms.
2. Estimating the Long-Run
Equilibrium Relationship
After testing for stationarity and finding that the data
series are stationary at first-difference, the next step is testing for cointegration
among I(1) variables in order to
determine their long-run relationship. The cointegration technique uses OLS to
estimate the cointegration regression with the following specification:
Equation (5.8) is regressed at level and the residuals are
tested for stationarity. The advantage of cointegration technique is that even
if the data series are not stationary at level, the regression at level is
still valid in order to determine the equilibrium (long-run) relationship among
the nonstationary time series variables because the superconsistency property
states that when the series are cointegrated, the OLS estimator of
nonstationary variables will approach its true values faster than the OLS with
stationary variables (Harris and Sollis 2003).
Estimating the long-term relationship by running an OLS
regression and obtaining the residuals for unit root testing is called the
first step in Engle-Granger two-step procedures. The null hypothesis for
residuals testing is 
against the
alternative that 
(Harris and Sollis 2003). The Augmented Dickey-Fuller (ADF)
test without constant and trend terms is used to test the stationarity of
residuals (Enders 1995). If we can reject the null hypothesis that the
residuals series contains a unit root, we can reject the hypothesis that the
variables are not cointegrated (Enders 1995). If the variables are
cointegrated, the coefficient estimates from the level regression represents
the long-run relationship (Shafik 1992).
against the
alternative that (Harris and Sollis 2003). The Augmented Dickey-Fuller (ADF)
test without constant and trend terms is used to test the stationarity of
residuals (Enders 1995). If we can reject the null hypothesis that the
residuals series contains a unit root, we can reject the hypothesis that the
variables are not cointegrated (Enders 1995). If the variables are
cointegrated, the coefficient estimates from the level regression represents
the long-run relationship (Shafik 1992).
Table 5.3 shows the results of estimating the long-run
relationship between private investment and the relevant regressors. The signs
of the coefficient are consistent with the theoretical expectations. The
accelerator effect is captured by the coefficient of the expected real GDP and
as expected, it has the correct sign for the all models.
The sign of the coefficient of real public investment is
positive in all models. This result is consistent with most of the empirical
result in the literature and the similar result by Ikhsan and Basri (1991) in
the context of Indonesia.
The relationship between private investment and financial
policy is measured by the banking credit to the private sector. The coefficient
of credit to the private sector shows a positive effect on private investment,
indicating that the availability of credit is important determinant of private
investment. This finding is in line with results found by Ramirez (1994) in the
context of Mexico.
Real exchange rate has a positive effect on real private
investment in Model (3), implying that exchange rate depreciation has increased
the competitiveness of the Indonesian export product. This result confirms a similar result by
Ikhsan and Basri (1991). Depreciation in the real exchange rate has a positive
impact on private investment because after the oil boom period ended, Indonesia
focused on a strategy to promote export-oriented investment (Hofman, Zhao and
Ishihara 2007). Indonesian exchange rate management policy is an effective tool
for maintaining competitiveness and encouraging private investment through
regular and major depreciations in 1978, 1983, 1986 and a free-floating in
1997.
The negative coefficient for the dummy variable indicates
that economic crisis had a negative impact on private investment.
To determine a long-term relationship between the variables,
the Augmented Dickey-Fuller (ADF) test is applied to the residuals. If the ADF statistic is smaller than the critical
values at the 1%, 5%, and 10% levels; the residuals are stationary. The ADF
statistics in Table 5.3 indicate the rejection of the null hypothesis of unit
root in the residuals, implying that the variables in the model are
cointegrated or have long-term relationships. This also means that the variation
in private investment can be explained by the relevant variables (Ramirez
1994).
The coefficient of the variables in the accelerator model in
logarithmic term has an interesting interpretation as elasticity (Ramirez
1994). From Model (3), the coefficient of the twice-lagged public investment
suggests that, ceteris paribus, increasing public investment by 10 percent will
increase private investment by 2.4 percent within two years.
The coefficients of adjustment (
) are calculated by subtracting the coefficient of lagged
private investment from 1. For Model (1) the coefficient of adjustment is .36
(1 - .64), meaning that 36% of the adjustment of actual investment to its
desired level occurs in one year.
) are calculated by subtracting the coefficient of lagged
private investment from 1. For Model (1) the coefficient of adjustment is .36
(1 - .64), meaning that 36% of the adjustment of actual investment to its
desired level occurs in one year.
Table 5.3 Determinants of Private
Investment: Level Regressions – Public Investment
Dependent
Variable: Real Private Investment
|
|||
Independent
Variables:
|
(1)
|
(2)
|
(3)
|
Constant
|
-0.781
(-1.795)
|
-0.717
(-2.061)
|
-0.375
(-0.999)
|
Expected
real GDP*
|
1.681
(3.391)
|
1.518
(3.685)
|
1.555
(3.608)
|
Lagged
real private investment
|
0.639
(4.179)
|
0.433
(4.303)
|
0.305
(2.260)
|
Real
public investment
|
0.056
(0.619)
|
-
|
-
|
Lagged
real public investment
|
-
|
0.206
(3.646)
|
-
|
Twice-lagged
real public investment
|
-
|
-
|
0.241
(3.574)
|
Real
credit to private sector
|
0.184
(2.244)
|
0.248
(4.021)
|
0.309
(4.180)
|
Real
exchange rate
|
0.116
(1.652)
|
0.124
(2.223)
|
0.124
(2.060)
|
Dummy
for economic crisis years
|
-0.218
(-2.735)
|
-0.167
(-2.505)
|
-0.133
(-1.862)
|
Observations
|
36
|
36
|
35
|
Adjusted
R2
|
0.98
|
0.98
|
0.98
|
DW
|
1.94
|
1.57
|
1.96
|
ADF
statistics for residuals (1% significance level)
|
-5.669
|
-5.015
|
-5.671
|
Coefficients
of adjustment
|
0.361
|
0.567
|
0.695
|
Notes:
-
Figures in parentheses are t-statistics.
-
* Expected real GDP = 
where depreciation
rate (
) is chosen to be 5 percent. 
is the predicted values obtained from estimating a
second-order autoregressive.
where depreciation
rate () is chosen to be 5 percent. is the predicted values obtained from estimating a
second-order autoregressive.
-
All variables are expressed in natural
logarithms.
3.
Estimating the Error Correction Model
Table 5.3 shows that the relevant variables in explaining the
variation in private investment are cointegrated. So, the residuals from the
equilibrium regression can be used to estimate the error correction model as
follows:
(5.9)
where 
is change in private
investment, 
is change in expected
real GDP 
, 
is change in public
investment, 
is change in the
availability of credit, 
is change in real
exchange rate, 
is the error
correction term which is the residuals from the long-run cointegration
regression, and 
is an error term. The
coefficient of the error correction term (
) is the disequilibrium error correction coefficient which
denotes the long-run speed of adjustment. If the relevant regressors are
cointegrated, the coefficient of the error correction term () should be negative and significantly different from zero
(Enders 1995). It measures the role of disequilibrium in explaining the
short-run movements in private investment. This being negative is a necessary
condition for the private investment to converge to the long-run equilibrium
relationship (Enders 1995).
is change in private
investment, is change in expected
real GDP , is change in public
investment, is change in the
availability of credit, is change in real
exchange rate, is the error
correction term which is the residuals from the long-run cointegration
regression, and is an error term. The
coefficient of the error correction term () is the disequilibrium error correction coefficient which
denotes the long-run speed of adjustment. If the relevant regressors are
cointegrated, the coefficient of the error correction term () should be negative and significantly different from zero
(Enders 1995). It measures the role of disequilibrium in explaining the
short-run movements in private investment. This being negative is a necessary
condition for the private investment to converge to the long-run equilibrium
relationship (Enders 1995).
Table 5.4 shows the error correction model of the private
investment determinants. The error correction term is statistically significant
in all models, signifying that the error correction model specification is
valid and the acceptance of the cointegration hypothesis. The value of the
error correction term .588 in Model 1 indicates that the adjustment in a
particular year will be .588 of any deviation from the equilibrium (long-run
relationship level). For example, if private investment is one percentage point
lower (higher) than the equilibrium, private investment will increase
(decrease) by .588 percentage points in the next year toward the equilibrium.
In all models, the coefficient for changes in expected real
GDP as an indicator of future demand, the changes in real credit to private
sector, and the changes in real exchange rate are positive and statistically
significance. Thus, in the short term, these variables have a positive impact
on private investment.
In Model 1 public investment crowds out private investment
with statistical significance. This result is consistent with similar findings
by Ikhsan and Basri (1991) and Everhart and Sumlinski (2001). A possible explanation
is that changes in public investment create inflation in the short term.
Increasing inflation raises the real interest rate which in turn decreases
investment. Other reasons for the crowding out effect are the existence of
inefficiency in public investment which made poor quality public investment, and
the existence of state-owned enterprises which discourage the private sector to
enter into a certain industry.
The residuals of the error correction model also pass the
diagnostic test of serial correlation, heteroskedasticity and normality (Table
5.5). The Breusch-Godfrey serial correlation the LM test show that LM statistic
(observation times R-squared) cannot
reject the null hypothesis of no serial correlation. The White
heteroskedasticity test show that the statistic (observation times R-squared) cannot reject the null
hypothesis of no heteroskedasticity. The Jarque-Bera statistic for testing
normality shows that the residuals are normally distributed.
Table 5.4 Determinants of Private
Investment: Error Correction Model
Dependent
Variable: ΔReal Private Investment
|
|||
Independent
Variables:
|
(1)
|
(2)
|
(3)
|
Constant
|
-0.004
(-0.232)
|
-0.036
(-2.272)**
|
-0.032
(-1.670)
|
Δ
expected real GDP
|
1.367
(5.563)***
|
1.148
(4.670)***
|
1.538
(4.284)***
|
ΔReal
public investment
|
-0.245
(-3.847)***
|
-
|
-
|
ΔLagged
real public investment
|
-
|
0.276
(4.424)***
|
-
|
ΔTwice-lagged
real public investment
|
-
|
-
|
-0.049
(-0.524)
|
ΔReal
credit to private sector
|
0.235
(4.197)***
|
0.287
(5.287)***
|
0.320
(4.673)***
|
ΔReal
exchange rate
|
0.119
(2.159)**
|
0.134
(2.248)**
|
0.107
(1.355)
|
Error
correction term
|
-0.588
(-4.395)***
|
-0.557
(-3.080)***
|
-0.678
(-2.968)***
|
Observations
|
32
|
33
|
32
|
Adjusted
R2
|
0.86
|
0.83
|
0.78
|
DW
|
1.96
|
1.76
|
1.80
|
Notes:
-
Figures in parentheses are t-statistics
-
*** 1% significance level. ** 5% significance
level. * 10% significance level.
-
Note: The error correction model is estimated
using OLS and the Cochrane-Orcutt iterative procedure (CORC) to correct any
serial correlation problem.
-
All variables are expressed in natural
logarithms.
Table 5.5 Residual Tests
Residual
tests
(Figures
in parentheses are p-value)
|
Model
1
|
Model
2
|
Model
3
|
Serial
correlation LM test
|
4.361 (0.113)
|
1.819 (0.403)
|
0.730 (0.694)
|
White
heteroskedasticity test
|
15.74 (0.110)
|
8.59 (0.571)
|
12.12 (0.277)
|
Jarque-Bera
|
0.973 (0.615)
|
0.803 (0.669)
|
0.412 (0.814)
|
The results that contemporaneous public investment has a
negative and significant impact on private investment deserve a special
explanation. It seems that Indonesia faces crowding out for the additional
spending of public investment. A possible explanation for this is the
traditional financial crowding out argument, namely, government projects
competing with private projects for limited financial resources the economy can
offer. Government projects may have a
positive supply-side effect, which may in turn stimulate private investment in
the end. But this effect takes time before it is felt by the private sector. In
this regard, it is interesting that the estimated coefficient for lagged public
investment is positive.
The existence of state-owned enterprises (SOE) which may
discourage the private sector from entering into a certain industry could be
another reason why public investment has a negative impact on private investment. As of 2001, there were 162 SOEs. Of these 124
were in competitive industries such as banking, construction, plantations and
trade; 12 held a monopoly from the government such as the airport service
authority; and 26 were in sectors that were mixed competitive, monopolistic or
had a public service-oriented market structure such as energy, transportation,
and telecommunications (Prasetiantono 2004). The private sector cannot compete
with SOEs because the SOEs usually obtain many privileges in economic resources
from the government such as to do with raw materials, the market or financial
resources (being bailed out when facing financial difficulties).
The privatization policy instituted by the government in the
early 1990s is still controversial and not completely successful even today.
The proponents of privatization argue that SOEs that incurred losses or lower
profitability should be privatized. The opponents argue that SOEs should not be
privatized because they play a strategic role as agents of development and are
operated as a public service (Prasetiantono 2004). As of June 2005, only 17
SOEs have been privatized (McLeod 2005). So the effect of SOEs on the
Indonesian private sector is still significant albeit they are often
inefficient, unprofitable and corrupt.
Regarding the effect of corruption on public investment, a
study by Tanzi and Davoodi (1997) shows that corruption reduces the
productivity and the quality of public investment. According to Everhart and Sumlinski (2001),
the crowding out effect of public investment is more influential when
corruption exists because corruption raise the level of public investment but
lower the quality of public investment, and based on their empirical result;
the low quality of public investment has a negative impact on private
investment. The sources of the problems such as road congestion and electric
power loss are usually the poor quality of public investment which reduces the
productivity of the private sector and increases the costs of the private sector.
It is no surprise that the poor quality of public investment
is caused by corruption. Corruption in Indonesia is still widespread. Many
international organizations rank Indonesia as one of the most corrupt countries
in the world. For example, Transparency International ranked Indonesia at 137th
in corruption perceptions index 2005 with score only 2.2 (10 = highly clean, 0
= highly corrupt) (Transparency International 2006). According to the late
Indonesian prominent economist Professor Sumitro Djoyohadikusumo, 30 percent of
the state budget was lost annually because of corruption (The Jakarta Post, January 27, 2003).
Even though public investment crowd out private investment
contemporaneously, and their lagged values generally have positive effects on
private investment. One can argue that any public investment tends to have
crowding out effects during the same period because of financial crowding out,
but after the public facilities are completed and start to provide services,
they encourage private sector investment. For example, public investment in
roads, power, or transportation may reduce the cost of production and thereby
encourage the private sector to invest after the construction is finished.
Then, the contemporaneous effect may be low or even negative, but over time,
the positive effect of public investment may emerge.
A possible reason for the unexpected and statistically
insignificant, sign of coefficients for twice-lagged public investment is
likely to be the small sample size. Box, Jenkins, and Reinsel (1994) recommend
50 and preferably 100 observations or more to be used for model building. This
study uses only 37 observations, which may weaken the significance of its
estimations.
The estimated coefficient for the error correction term is
negative and highly significant in all models, verifying the existence of a
stable long-run relationship between private investment and the explanatory
variables. A negative coefficient means that when there is a discrepancy
between actual and long-run relationship, there is an adjustment toward the
long-run equilibrium in later periods to eliminate this discrepancy. If the private investment is
higher (lower) than that of the long-run equilibrium level, the error
correction term will adjust it back to the equilibrium. The absolute values of
the error correction term for all models are high, greater than .50, meaning that
the adjustment is fast. For example, in Model (1) that includes the current
public investment variable, around 59% of any deviation from the long run
equilibrium is adjusted in one year. For Model 3 which incorporates
twice-lagged public investment, the adjustment is 68% in one year.
VI. CONCLUSIONS AND
POLICY RECOMMENDATIONS
1. Conclusions
Investment is important for economic growth and poverty
reduction. Together with eradicating corruption and reducing poverty, increasing
private sector investment is top priority among the government’s policy
objectives. So it is important for policy makers to understand how private investment
responds to changes in policies. Using the flexible accelerator model as a
framework for analyzing the variables that systematically affect private
investment, this study considered the roles of certain government policy
instruments such as public investment, as well as those of exchange rate and
the availability of credit to the private sector.
The econometric approach addressed the problem of spurious
regression by testing the variables for stationarity. The Engle-Granger
cointegration test was used to determine the long-term relationship between
variables. Lastly, this study estimated an error correction model to capture
the short-run dynamics of private investment.
The Augmented Dickey-Fuller and the Phillips-Perron tests for
unit roots indicated that all the variables were integrated of order one, I(1). The Augmented Dickey-Fuller tests
on residuals of the level regressions suggested the rejection of a unit root in
the residuals. Finally, the coefficient for the error correction terms is
statistically significance and in the correct sign, implying that error
correction toward the long-run equilibrium is indeed taking place.
The results obtained from the error correction model showed
that current public investment had a negative and significant effect on private
investment during the 1970-2006 periods. However, when public investment was
lagged one period it showed a positive and significant impact on private
investment. Only lagged
public investment result confirmed the importance of physical capital for private
investment. There are three possible reasons for the above results. First, the
quality of public investment may be poor, perhaps due to corruption. Second,
the dominance in certain sectors of state-owned enterprises may be discouraging
private sector investment. Finally, the small sample can be the cause of low
significance.
The expected output variable which represents the accelerator
effect, and credit to private sector and the real exchange rate as control
variables, showed a positive and significant effect on private investment.
2. Limitations and
Further Studies
The series for the public investment was not disaggregated
into infrastructure and noninfrastructure, health and education. Ideally, data
on similar spending should be disaggregated. It is expected that if disaggregated
data was used, the regression would produce a better result. So it is recommended
to use disaggregated data in the future studies.
Another limitation is that the number of observations is
relatively small. This study used annual data 1970-2006. Even though this is
quite long time span, the data points provides only 37 observations, less than
the 50 suggested by Box, Jenkins, and Reinsel (1994). Because obtaining annual
data for 50 years is very difficult, future studies should use higher frequency
data such as quarterly data.
3. Policy Recommendations
Based on the results from Part 5, we make the following
policy recommendations which are not limited to the variables that determine
private investment but also include relevant factors that affect the behavior
of private investment in Indonesia:
1. Government
should be concerned with not only the level of public expenditure but also its
composition. For example government expenditure should be focused on
infrastructure, education and health. Unproductive expenditures such as fuel
and electricity subsidies should be reduced gradually. Reducing the subsidies
cannot be achieved without strong political will from the government.
2. If
the crowding out effect of the public investment is due to the poor quality of
spending, the government should intensify its efforts to minimize the cause of
the poor quality. Politically intervention in the procurement process of public
spending should be prevented. The government should also avoid initiating white
elephant projects which may be politically appealing but are not based on
feasibility studies under cost and benefit analysis. The corruption eradication
effort should also be intensified as corruption reduces the productivity and
quality of public investment.
3. To
minimize the crowding out effect from state-owned enterprises, government
should promote privatization in a transparent way. The privatization process in
Indonesia tends to create controversy and faces many obstacles from the
management, the employees, local governments, and public. In order to reduce
controversy and obstacles, there must be transparency in the privatization
process and government commitment to resolve any problems that may arise during
and after privatization. For example, during privatization process government
should accommodate the interests of the management, employees, local
governments and public. After privatization, government should develop credible
regulations, so that public still has access to the services at reasonable
price.
4. The
empirical results show that increasing bank credit is associated with higher
private investment. Thus, easy and adequate access to credit is crucial for the
development of the private sector.
REFERENCES
Biro Pusat Statistik (Central
Statistic Bureau) (1988), “Analisa
Investasi,” (Investment Analysis) Jakarta: BPS.
Blejer, Mario and Khan, Mohsin S. (1984) “Government Policy and Private
Investment in Developing Countries” IMF
Staff Paper, vol. 31, no. 2, June.
Box, George E.P.; Jenkins, Gwilym M.; and Reinsel,
Gregory C. (1994), “Time Series Analysis:
Forecasting and Control,” Third Edition, New Jersey: Prentice Hall.
Chenery, Hollis B. (1952), “Overcapacity and the
Acceleration Principle,” Econometrica
20.
Chhibber, Ajay and Dailami, Mansoor (1990), “Fiscal Policy and Private Investment in
Developing Countries: Recent Evidence on Key Selected Issues,” Working
Papers, the World Bank.
Chhibber, Ajay and van Wijnbergen, Sweder (1988), “Public Policy and Private Investment in Turkey,”
Working Paper Series 120, Washington DC: The World Bank.
Chhibber, Ajay; Dailami,
Mansoor; and Shafik, Nemat (1992), “Reviving
Private Investment in Developing Countries: Empirical Studies and Policy
Lessons,” Amsterdam: North-Holland.
Chirinko, Robert S. (1993), “Business Fixed
Investment Spending: Modeling Strategies, Empirical Results, and Policy
Implications,” Journal of Economic
Literature 31.
Clark, John Maurice (1917), “Business Acceleration
and the Law of Demand,” Journal of
Political Economy, 25.
Cruz, Bruno de Oliveira and Teixeira, Joanilio R.
(1999), “The Impact of Public Investment on Private Investmen in Brazil,
1947-1990,” CEPAL Review, 67, April.
Diebold, Francis X. (2007), “Elements of Forecasting,” Fourth Edition, Ohio: Thomson
South-Western.
Erden, Lutfi and Holcombe, Randall G. (2005) “The Effects of Public
Investment on Private Investment in Developing Economies” Public Finance Review, vol. 33, no. 5, September.
Enders, Walter (1995), “Applied Economic Time Series,” New York: John Wiley & Sons.
Engle, R.F. and Granger, C.W.J. (1987), “Cointegration
and Error Correction: Representation, Estimation, and Testing,” Econometrica, 55.
Everhart, Stephen S. and Sumlinski, Mariusz A.
(2001), “Trends in Private Investment in
Developing Countries Statistics for 1970-2000 and the Impact on Private
Investment of Corruption and the Quality of Public Investment,” Discussion
Paper No. 44, International Finance Corporation, Washington DC: the World Bank.
Ghura, Dhaneshwar and Goodwin, Barry (2000), “Determinants
of Private Investment: A Cross-Regional Empirical Investigation,” Applied Economics, 32.
Goodwin, R.M. (1951), “The Non-linear Acceleration
and the Persistence of Business Cycles,” Econometrica,
Vol. 19.
Greene, J. and Villanueva, D. (1991), “Private
Investment in Developing Countries: An Empirical Analysis,” IMF Staff Papers, No. 1, March, Vol. 38.
Hall, Robert E. and Jorgensen, Dale W. (1967), “Tax
Policy and Investment Behavior,” American
Economic Review 57.
Harris, Richard and Sollis, Robert (2003), “Applied Time Series Modelling and
Forecasting,” England: John Wiley & Sons.
Ikhsan, Mohamad and Basri, M. Chatib (1991). “Investasi
Swasta dan Pemerintah; Substitusi atau Komplementer?: Kasus Indonesia (Private
and Government Investment: Substitution or Complementary?: Indonesian Case)” Ekonomi dan Keuangan Indonesia, Vol. 39,
No. 4.
Jorgensen, Dale W. (1963), “Capital Theory and
Investment Behavior,” American Economic
Review 53.
Jorgensen, Dale W. (1971), “Econometric Studies of
Investment Behavior: A Survey,” Journal
of Economic Literature 9.
Jorgensen, Dale W. and Siebert, Calvin D. (1968), “A
Comparison of Alternative Theories of Corporate Investment Behavior,” American Economic Review 58.
McKinnon, Ronald (1973), “Money and Capital in Economic Development,” Washington DC: The
Brooking Institution.
Naqvi, Naveed H. (2002), “Crowding-in or
Crowding-out? Modelling the Relationship between Public and Private Fixed
Capital Formation Using Co-integration Analysis: The Case of Pakistan
1964-2000,” The Pakistan Development
Review, 41:3 Autumn.
Prasetiantono, Tony (2004), “Political Economy of
Privatization of State-Owned Enterprises in Indonesia,”in Basri, M. Chatib and
van der Eng, Pierre (2004), “Business in
Indonesia: Old Problems and New Challenges,” Singapore: Institute of
Southeast Asian Studies.
Ramirez, Miquel D. (1994), “Public and Private
Investment in Mexico, 1950-90: An Empirical Analysis,” Southern Economic Journal, July, Vol. 61, No. 1.
Ribeiro, Marcio Bruno and Teixeira, Joanilio Rodolpho
(2001). “An Econometric Analysis of Private-Sector Investment in Brazil,” CEPAL
Review, 74.
Sakr, Khaled (1993), “Determinants of Private Investment in Pakistan,” Working Paper
WP/93/30, Washington DC: International Monetary Fund.
Saxton, Jim (2002), “Federal Individual Income Taxes and Investment: Examining The Empirical
Evidence,” Washington: Joint Economic Committee.
Serven, L. and Solimano, A. (1989), “Private
Investment and Macroeconomic Adjustment: An Overview,” PPR Working Paper Series WPS 339, Washington DC: The World Bank.
Serven, Luis (1996), “Does Public Capital Crowd Out Private Capital?” Policy Research
Working Paper 1613, Washington DC: World Bank.
Shafik, Nemat (1990), “Modeling Investment Behavior in Developing Countries: An Application to
Egypt,” Working Paper Series WPS 452, Washington DC: The World Bank.
Shaw, E. (1973), “Financial
Deepening in Economic Development,” New York: Oxford University Press.
Sundararajan, V., and Thakur, S. (1980), “Public
Investment, Crowding Out, and Growth: A Dynamic Model Applied to India and
Korea,” IMF Staff Papers, Vol. 27,
December.
Tanzi, Vito and
Davoodi, Hamid (1997), “Corruption, Public Investment, and Growth” IMF Working Paper WP/37/139. Washington
DC.
The Jakarta Post, January 27, 2003, Editorial, Embezzling foreign
loans.
Tobin, James (1969), “A General Approach to Monetary
Theory,” Journal of Money, Credit, and
Banking 1.
Transparency International (2006), “Global Corruption Report 2006,” Berlin:
Transparency International.
Valadkhani, Abbas (2004), “What Determine Private
Investment in Iran?” International
Journal of Social Economics, Vol. 31 No 5/6.
Wai, U Tun and Wong, Chorng-huey (1992), “Determinants
of Private Investment in Developing Countries,” The Journal of Development Studies, Vol. 19, No. 1, October.
Widarjono, Agus (2005), “Ekonometrika: Teori dan Aplikasi untuk Ekonomi dan Bisnis,”
(Econometrics: Theory and Application for Economics and Business), Yogyakarta:
Ekonisia.
Wooldridge, Jeffrey M. (2006), “Introductory Econometrics: A Modern Approach,” Third Edition, Ohio:
Thomson South-Western.
World Bank (2007) Spending
for Development: Making the Most of Indonesia’s New
Opportunities, Indonesia Public Expenditure Review 2007. Washington DC: World Bank.
[1] The data
limitations in developing countries include the investment data in gross terms,
no published capital stock series, and no appropriate measures of return on
investment (Ramirez (1994) and Ghura-Goodwin (2000)).
[2] In order to get
the real value, the nominal value is deflated by CPI (2000 = 100), except for
real GDP in which nominal GDP is deflated by the GDP deflator (2000 = 100) and
gross fixed capital formation is already in 2000 constant prices from the
Central Statistic Agency.
No comments:
Post a Comment