The ”Battle” of Fiscal Instruments: Evidence from

Indonesia
Alfan Mansura,b,∗, Ardi Sugiyartob , Pipin Prasetyonob , Muhammad Afdi
Nizarb
a
University of Helsinki, Arkadiankatu 7, Helsinki, 00100, , Finland
b
Fiscal Policy Agency, Ministry of Finance, Jl. Dr. Wahidin
1, Jakarta, 10710, , Indonesia

Abstract

This study examines the macroeconomic outcomes of different fiscal instru-

ments using Indonesian data from 1969 to 2020, including the Covid-19 crisis
period. We base our analysis on a structural vector autoregression (SVAR)
model by extracting the impulse responses and historical decomposition to
simulate the economic performance with and without a fiscal policy. We
find asymmetrical effects between spending and tax where the latter is found
more powerful. We also find more adverse impacts of increased VAT on
consumption than reduced-current-expenditure and a higher positive effect
of lowered-income-tax on investment in the short term, but longer-lasting
impacts of raised-capital-expenditure.
Keywords: fiscal multiplier, tax, spending, SVAR

∗
Corresponding author
Email address: [Link]@[Link] (Alfan Mansur)

Preprint submitted to Journal of Policy Modeling May 5, 2022

Electronic copy available at: [Link]

1. Introduction

Fiscal policymakers always face challenges in formulating an optimal pol-

icy choice that benefits the economy the most. Theoretically, the goal is to
stabilize the economic performance with available resources given any con-
straints. Two policy options for fiscal authority are to optimize the tax tool
and the spending tool. The roles of the tax and spending tools are even
more prominent during a crisis period to soften the effects on the economy.
Spending expects to be higher, while taxes are to be lower. Some evidence
shows that fiscal multipliers are significantly larger during recessions or slack
times (Peren Arin et al. 2015; Canzoneri et al. 2016; Lee et al. 2020).
As important as learning from advanced economies, learning from other
developing and emerging market (EM) economies is also beneficial. As with
other EM economies, Indonesia also features the same commonalities with
other developing and EM economies. Indonesia and the other six largest EM
economies, i. e., China, India, Brazil, Russia, Mexico, and Turkey, account
for more than half of global output growth during 2010-2015 (Huidrom et al.
2020). With a long history of its fiscal policy, there are abundant informa-
tion and policy lessons that other EM economies may gain from Indonesia’s
experience.
Indonesia has experienced several crises, from the Asian Financial Crisis
1998 to the Covid-19 outbreak 2020. During the latter, Indonesia’s economy
contracted by 2.07%, while government spending increased by 12.15% and
tax decreased by 16.68%. This figure theoretically indicates that the tax tool
is more optimized than the spending tool. Which one is more optimum for
the economy is still understudied from the scholar’s point of view.

Electronic copy available at: [Link]

 One study assessing the effectiveness of fiscal policy in Indonesia is Yien
et al. (2019) using the data of 1970-2015. Based on their Toda-Yamamoto
causality test results, both government expenditure and tax insignificantly
cause gross domestic product per capita (GDP per capita). Nevertheless, the
latter is found significant to cause the expenditure and tax implying for pas-
sive fiscal policy. Consequently, this is less helpful for the fiscal policymaker
since we believe that a fiscal policymaker should pursue an active approach.
Another study by Tang et al. (2013) reveal a statistically insignificant effect
of a spending shock to GDP and significantly positive response of GDP to
a tax shock. Moreover, they find a negative impact fiscal multiplier from
a spending shock, while positive impact fiscal multiplier from a tax shock
which literature regards this as expansionary fiscal contractions. These re-
sults raise our concern since if that is the case, then we can be better off by
doing no spending and raising tax as high as possible until reaching some
point. We know that this cannot be true.
Hence, in this research, we propose several things for a more compre-
hensive study of the effects of fiscal policy shocks. First, we argue that an
expansionary fiscal contraction can be the case that there is the same shock
affecting both GDP and tax at the same time. Hence, we here propose to
adjust the tax series cyclically. Second, the analysis should go into more
detail at least, each of spending and tax should go into more detail based on
their types. We believe that different types of spending or tax will generate
unequal effects on the economy depending on their transmission channels.
For this argument, we provide an analysis of the impulse response function
to examine the impacts of income tax shock on investment and VAT shock on

Electronic copy available at: [Link]

consumption. We here also provide a counterfactual analysis of the economic
performance, with the presence and absence of fiscal in the economy, which
subsequently feeds us in evaluating the specific periods where we perform
well or not.
The remainder of this paper is structured as follows. Section 2 discusses
the theoretical background of fiscal transmission to the economy, and Sec-
tion 3 portrays the fiscal policy development in Indonesia. Section 4 then
describes the methodology subsequently the data used in this paper. Then,
the subsequent two sections discuss the impacts of government spending and
tax on the Indonesian economy and the policy implications. In the policy
implication section, we also cast policy simulation of the effects of reducing
current expenditure versus raising VAT on consumption and increasing capi-
tal expenditure versus raising the income tax on investment. The last section
finally concludes.

2. Transmission mechanism of fiscal policy and multipliers: a lit-

erature review

2.1. Transmission mechanism

Theoretically, fiscal policy through spending and tax aims to correct tem-
porary disequilibria in the economy, stimulate economic growth, or redis-
tribute income. To achieve these goals, three main functions are formu-
lated, namely: (i) allocation, (ii) distribution, and (iii) stabilization functions
(Musgrave 1959). For allocation, Government may intervene either directly
through direct purchases of goods and services or indirectly through taxes
and subsidies. Then for the latter two, Government can provide public goods

Electronic copy available at: [Link]

and alike to attain full employment and stable prices. As far as efficiency
is concerned, more specific targets are efficient use of resources, equal distri-
bution of income, and mitigated business cycle fluctuations (Tanzi and Zee
1997). All those objectives are often pursued via manipulation of aggregate
demand.
Three different perspectives to find out how the transmission mechanism
of fiscal policy takes place in the economy include (i) demand-side, (ii) supply-
side, and (iii) institutional aspects (Hemming et al. 2002). The demand-side
effects of the fiscal policy refer to two foremost mainstreams that follow the
Keynesian and Non-Keynesian approaches. In the Keynesian model, fiscal
policy affects output via aggregate demand. A fiscal expansion, for example,
may induce households’ disposable income and consumption either directly
through spending or indirectly through tax cuts or transfer payments.
Meanwhile, the rational expectation theory influences the Non-Keynesian
model, especially the Neo-Classical model. Consumers are forward-looking,
meaning they know the government’s budgetary constraints on the aggregate
demand. Temporary fiscal expansion that has no long-term effect will not
affect expectations, but permanent fiscal expansion will. In turn, the latter
can encourage higher interest rates and a greater appreciation of exchange
rates, thereby inducing crowding out (Hemming et al. 2002; Hebous 2011).
On the other hand, a fiscal contraction is to lower interest rates, which en-
courage investment and reduce the negative output effect of the contraction
and even turn it into a positive impact (Alesina and Ardagna 1998).
The supply-side effects of fiscal policy, the emphasis of the Non-Keynesian
model, usually focus on the impact of taxation on labor income. Since labor

Electronic copy available at: [Link]

supply and capital taxes are affected, savings and investment change. Worker
consumption will also fall after a tax increase, and to maintain consumption,
the labor supply needs to increase, and vice versa, so do the labor’s spending
and its capital productivity (Hemming et al. 2002). These demand and
supply sides of fiscal policy impact are essential indicators, among others,
to measure the effectiveness of the fiscal policy transmission process, which
usually uses a fiscal multiplier.

2.2. Fiscal multipliers

The significant role of fiscal policy in the economy, especially expansion-

ary fiscal policy during the economic downturns, has sparked the discourse
of measuring fiscal multipliers by scholars. In general, the fiscal multiplier
estimates the final change in real national income resulting from exogenous
changes in government spending or government revenue (Batini et al. 2014).
The concept of the fiscal multiplier is rooted in the importance of calculating
the effect of a certain amount of change in tax expenditure or revenue on
the level of GDP. Understanding the size of this multiplier means that the
fiscal policymaking process will be more effective in determining how much
change is needed in the government budget to respond to economic condi-
tions. In addition, an accurate calculation of the fiscal multiplier will benefit
fiscal authority by better forecasting economic growth, which in turn spurs
its credibility.
In measuring fiscal multipliers, the idea of whether fiscal policy affects
the economy through demand-side or supply-side would lead to different
approaches and ways of thinking. First, demand-side or Keynesian multi-
plier stresses the role of government spending in spurring aggregate demand,

Electronic copy available at: [Link]

rather than employing tax policies since it assumes that the increase in de-
mand caused by spending increases is higher than that is caused by tax cuts
(Hemming et al. 2002). Batini et al. (2014) argue that during the implemen-
tation of the tax-cuts policy, households tend to save a noticeable part of
additional income rather than spend it. Therefore, tax cuts are considered
less influential in stimulating the economy than spending increases. In a sim-
ple Keynesian model, the transmission of government spending on aggregate
demand is heavily affected by the responsiveness of households’ consumption
to their current income or marginal propensity to consume (MPC). However,
since this transmission involves and induces investment in the private sector,
the Keynesian multiplier is, therefore, greater than 1 (Hemming et al. 2002).
On the other hand, the demand-side Neo-Classical model usually per-
ceives the fiscal multiplier to be lower than 1. Several factors, i.e., current
income, consumers’ rational expectation, and their forward-looking estimate
about the government’s budget constraints, affect households’ consumption
decisions, subsequently inducing a lower fiscal multiplier (Hemming et al.
2002). Such factors undermine current consumption decisions because house-
holds and firms adjust their spending decisions. They would anticipate higher
taxes or higher interest rates in the future by increasing current savings and
reducing current consumption. In the case of full Ricardian equivalence, for
instance, fiscal multipliers are even perceived to be zero since expectations
about higher tax rates in the future lead people to save the current additional
income rather than spend it. Therefore, the effect of tax cuts on the current
period on aggregate demand is theoretically zero (Hemming et al. 2002).
Unlike the demand-side, the supporters of supply-side effects of fiscal

Electronic copy available at: [Link]

policy measure the multipliers by primarily estimating the transmission of
tax policy changes to the changes in labor income, saving, and investment,
which sequentially affect aggregate supply. Lower business taxes contemplate
improving capital investment spending, while lower-income taxes deliberate
improving working incentives. In addition, the decision to cut consumption
taxes might also indirectly affect labor supply or firms’ investment (Hemming
et al. 2002). However, the role of government spending in stimulating the
supply side cannot be neglected. Any government spending related to higher
labor productivity and capital accumulation, such as human capital spending
in health and education (Shafuda and De 2020) and public infrastructure
spending (Murphy et al. 1989), are responsible for higher fiscal multipliers.

3. Fiscal policy development in Indonesia

3.1. Public finance management

At the beginning of the “New Order” era, an era established by the sec-
ond President Soeharto in the late 1960s, Indonesia’s fiscal policy focused
on economic recovery to tackle hyperinflation and severe debt burden. Fis-
cal policy directed to achieve an internal balance condition where domestic
revenue could finance all operational expenditure while capital expenditure
would be financed by foreign grants or loans (Seda 2009). Although fiscal
discipline was one of the features of a balanced budget, the government still
wanted to accelerate economic development progress, so the government set
a dynamic balance budget instead of a static budget. It implied that revenue
would increase gradually, so it could accumulate public savings to finance
the economic development.

Electronic copy available at: [Link]

 After the fall of the New Order regime in 1998, the state budget started to
adopt a modern budget system proposed by IMF. In the new budget system,
the government adopted a new budgeting format that disclosed all revenue,
spending, and financing in a unified I-account format. In this budgeting
format, the budget transformed into a unified budget which was necessary
to avoid duplication and to bring all off-budget activities into the budget.
During 2003-2004, the government issued 3 public finance laws to promote ef-
ficient, effective, transparent, and accountable management of public finance
(Ministry of Finance of the Republic of Indonesia 2005). Since then, Law No.
23/2003 has provided a fundamental guideline to maintain fiscal deficit by
applying the Maastricht Treaty, which sets a limit for the national deficits
and borrowing by 3 percent of GDP and 60 percent of GDP respectively
(Morris et al. 2006).
The public finance management in Indonesia has also experienced a dra-
matic change as the decentralization program has been performed since 2001.
Transfer of authority to the regional governments (district and province gov-
ernment) in several affairs has been followed by massive fiscal transfer to
the regions. Transfer to the region accounts for about one-third of the total
central government budget (Abdurohman 2013). Revenue sharing transfer,
general allocation fund, and special autonomy fund become a mandatory ex-
penditure of central government that should be allocated to regions according
to some specific fiscal rules. These funds are determined based on a specific
portion of the central government revenues.

Electronic copy available at: [Link]

3.2. Tax reform and spending policy in Indonesia

To enhance and maintain government revenue stability, a series of tax re-

forms were conducted between 1983 and 2000, indicated by several revisions
of tax law regulations. The first tax reform was in 1983, intended to reduce
the dependency on natural resources-based revenue and improve tax com-
pliance. To increase tax revenue and boost competitiveness in exports, the
government abolished sales tax and introduced Value Added Tax (VAT). In
addition, the tariff of the highest bracket of personal income tax lowered from
45% to 35%. Moreover, to improve collection performance, a self-assessment
approach in the tax system fundamentally marked tax administration reform
in 1983.
The second tax reform in Indonesia was implemented in 2002 by modern-
izing the tax administration. Having aimed to increase their capacity and to
improve the effectiveness of their tax service as well as their tax collection,
the Directorate General of Tax (DGT) of Indonesia established many Large
Taxpayer Offices (LTOs). DGT continued to restructure and modernize its
regional tax offices gradually until the end of 2008 (Eka 2019). In terms of
tax policy, Income Tax Law 2008 has allowed changes in personal exemptions
providing tax relief and has introduced more lenient tax brackets (Wijayanto
and Vidyattama 2017). In 2009, VAT law was amended for the third time
to promote the simplicity of VAT and to regulate the VAT exemption.
Post Tax reform in the 2000s, tax revenue gradually improved and was
stable at 11% of GDP until 2014. To improve the tax revenue collection and
compliance, the Government launched a tax amnesty program in 2016-2017,
which then successfully generated additional revenues of more than Rp114

10

Electronic copy available at: [Link]

trillion. Further tax policy reform has been implemented as an integrative
part of the Omnibus Law of Job Creation issued in 2020. To ease business
and attract investment, Indonesia has offered several tax incentives including
a reduction in corporate income tax, tax exemption of domestic dividend
revenue, and a reduction of income tax tariff for bond interest revenue.
In terms of public expenditure, Indonesia has a relatively low expendi-
ture size compared to other countries. The low size expenditure has been
prolonged since the end of the commodity boom in the 2000s. Indonesia
maintained government expenditure of 17% of GDP on average during 2001-
2014 and well below 15% of GDP after 2014. One milestone in the Indonesian
spending policy was in 2014 when there was a big reform in energy subsidies.
It then allowed the Government to remove gasoline subsidies which created
additional fiscal space that was beneficial in improving capital spending.
Current spending, subsidies, and transfer to the regions are now the
biggest portion of government expenditure and for this reason, the alloca-
tion of capital spending in Indonesia is still relatively low. The declining
trend of revenue collection causes a lower government spending size. More-
over, mandatory spending on education and health coupled with interest
payments result in a limited space to finance new programs. It then creates
an adverse impact on budget flexibility and the efficiency of the use of public
resources (Bank 2020).

3.3. Fiscal policy in 3 crisis periods

Back more than two decades ago, during the Asian Financial Crisis (AFC)
1998, the government’s financial position had been adversely affected because
of the sharp decrease of GDP growth, currency depreciation, and skyrocket-

11

Electronic copy available at: [Link]

ing inflation rates. However, in this period, government revenue looked well
maintained because several types of revenue increased as a consequence of
the depreciation and hyperinflation. In literature, such a phenomenon was
known as the ‘seigniorage’ effect. On the expenditure side, the depreciation
and high-interest rate caused higher liabilities of interest payment. Energy
subsidy had increased significantly as a result of currency depreciation. In
addition, current expenditure was adjusted to compensate for the higher in-
flation.
To counter the crisis, fiscal policy in 1998 was guided towards more ex-
pansionary policy. The 1998 budget ran deficit to around Rp11 trillion (1.2
percent of GDP), altering from initially a surplus of around Rp4 trillion (0.6
percent of GDP) in 1997 (Abdurohman 2013). The capital spending was
maintained increasing to stimulate the domestic economy and at the same
time absorbing unemployment. An additional subsidy was introduced to
maintain the purchasing power of people to protect the poor.
After 1998, Indonesia’s economic performance showed progressive devel-
opment. From 2000 through 2008, economic growth was stable on average
of 5% and peaked at 6.3% in 2007. However, the Global Financial Crisis
(GFC) distorted this trend in 2008. In response to the crisis, in early 2009
the Indonesian government launched a countercyclical fiscal stimulus pack-
age (FSP) to stimulate the domestic economy. Fiscal policy in 2009 was set
more expansionary by the implementation of tax cuts, tax subsidies, and the
introduction of more government investment projects (Basri 2013). The key
difference between fiscal policy response to GFC 2009 and AFC 1998 is the
source of financing of the stimulus. In 2009, the fiscal stimulus package im-

12

Electronic copy available at: [Link]

plemented was mainly financed through domestic bonds, while fiscal stimulus
in 1998 was mainly financed by the higher government revenue and foreign
financing, both through program and project loans.
Between 2010 and 2019, the Indonesian economy had been stable with
an average growth of 5.5%. The commodity boom in 2010-2014 supported
the high economic growth and good performance of the government revenue
collection. Government spending’s size was stable in the range of 15-16% of
GDP, but in a declining trend because of the down-trending revenue, while
at the same time Government maintained the budget deficit below than 3%
of GDP.
The Government budget changed drastically when pandemic Covid 19
occurred in 2020. Government revenue shrank sharply as business activities
contracted while expenditure increased to cover all the costs to mitigate the
impact of the crisis. Overall, the fiscal policy during the pandemic of 2020
was set more countercyclical. It was indicated by the massive tax cut and
immense additional government expenditure, particularly to address health
issues. Budget deficit 2020 relaxed to exceed 3% of GDP. The Indonesian gov-
ernment introduced National Economic Recovery (PEN) Program as much
as Rp 695,2 trillion as a response to the Covid 19 crisis. It aimed to sup-
port low-income households, the health sector, enterprise and SMEs, and to
support bank liquidity (Fukuchi 2020). To finance the budget deficit, the
government issued more bonds and developed a burden-sharing scheme with
the central bank. In this scheme, Bank Indonesia was allowed to purchase
government bonds and to play a role as a standby buyer in the primary
market (Habir and Wardana 2020).

13

Electronic copy available at: [Link]

 AFC 1998 GFC 2009 Covid 19-20
Fiscal policy stance expansionary expansionary expansionary
Stimulus Labour intensive; Personal Income and Tax cut; Health-
infrastructure pro- Corporate Tax Sav- related spending;
gram ing; Infrastructure Social assistance
Spending spending; Subsidy;
Liquidity support to
banks
Revenue +40% -14% -13%
Expenditure +55% -5% +18%
Deficit -1.2% GDP 1.5% GDP -6.14% GDP

Table 1: Fiscal policy in 3 crises

4. Methodology and data

4.1. Model

Adopted from Blanchard and Perotti (2002), this paper applies a trivari-
ate Structural Vector Autoregressive (SVAR) model as follows:
4
X
zt = c + κt + Ai zt−i + B0−1 εt (1)
i=1

where zt is a vector of endogenous variables containing government spending

(gt ), government tax (tt ), and GDP (yt ). c is a constant and κt is a time trend.
εt contains [εgt , εtt , εos
t ] which are structural shocks of government spending,

tax, and other, respectively. We label the last shock as other shock since we
here focus on fiscal shocks and to label the last shock as a particular shock,

14

Electronic copy available at: [Link]

more information is [Link], Ai is matrix of lag coefficients, while
B0−1 is coefficient matrix of the structural error terms.
As for identification, gt in equation 1 is ordered first using qualitative
information that government spending is firstly decided before the decision
on tax. In certain periods, tax target is firstly decided before the spending
and this information will be used as a robustness check for the model. gt
and tt in equation 1 both enter the model as cyclically adjusted series to
remove collinearities between both, so that produced estimates are unbiased.
It is also intended to neutralised the tax shock from the common shock
affecting GDP. These cyclically adjusted series for log tax and log spending
are constructed by the following formulas:

tt,c.a. = tt − aty ∗ yt (2)

where the left hand side is the cyclically adjusted series of log tax obtained
from the elasticity of tax to GDP (aty ) multiplied by log GDP (yt ) subtracted
from the log tax (tt ) and

gt,c.a. = gt − agy ∗ yt (3)

where the left hand side is the cyclically adjusted series of log spending
obtained from the elasticity of spending to GDP (agy ) multiplied by log
GDP (yt ) subtracted from the log spending (gt ). The endogenous variables
are in log levels and estimating an SVAR model with series in levels is proved
consistent (Kilian and Lütkepohl 2017).
The estimated structural restriction in the uncorrelated error terms εt is

15

Electronic copy available at: [Link]

as follows:
   
bˆ bˆ bˆ 6.8520 0 0
 11 12 13   
Bˆ0−1 = bˆ21 bˆ22 bˆ23  = 1.3828 2.7874 1.8562 (4)
   
   
ˆ ˆ
b31 b32 b33ˆ 0.0013 −0.0003 0.0120

where in Blanchard and Perotti (2002) bˆ23 = 2.08. Here for Indonesia case,
it is 1.8562 reflecting the average of the elasticity of government tax with
respect to GDP over the 1969:II-2020:IV period.
Having checked eigenvalues of the matrix Ai above, the model satisfies
the required stability condition. For estimating the sub-models, equations (1)
and (4) are re-estimated with zt comprising capital expenditure, income tax,
investment and another specification of zt containing current expenditure,
VAT, and consumption. The former is to study the effects of the fiscal
instruments on investment and the latter is to study the effects of the fiscal
instruments on consumption.

4.2. Statistical check for identification

In this part, we test statistically for the zero restrictions of some ele-
ments of B0−1 matrix in equation (1) as for a robustness check. Let the
reduced-form errors be et where et = B0−1 εt as expressed in equation (1).
Adapted from Rigobon (2003) and Lanne and Lütkepohl (2008), suppose
there is an exogenous change in the covariance matrix of the reduced-form
errors at time T1 : E(et e′t ) = ′
P P
1 , t = 1, . . . , T1 and E(et et ) = 2, t =

T1 + 1, . . . , T . Then there exists a (KK) matrix B0−1 and a diagonal matrix

′
Λ = diag(λ1 , . . . , λK ), λi > 0, i = 1, . . . , K, such that 1 = B0−1 B0−1 and
P

−1 −1′
P
2 = B0 ΛB0 . The covariance matrix of the structural errors εt = B0 et

16

Electronic copy available at: [Link]

is given by 
IK , t = 1, ..., T1

E(εt ε′t ) = (5)
Λ, t = T1 + 1, ..., T.


Apart from changes of signs and permutations of the columns, B0−1 is

unique if the diagonal elements of Λ i.e. λ1 , . . . , λK are distinct (see the proof
e.g. in Appendix A of Rigobon (2003)). Only if this condition is satisfied,
the zero restrictions embodied in the B0−1 matrix can be statistically tested.
Since we have a long dataset from 1969: II to 2020:IV, we then have the
period of Asian Financial Crisis 1998 in between. By plotting the reduced-
form errors as depicted in Figure 1, we can see a clear slump during the
Asian Financial Crisis 1998 period particularly on e3t which corresponds to
the real GDP. Using this information, we can set T1 = 1997 : IV as the point
of structural break and since our model is trivariate, we set K = 3, hence we
need the condition of λ1 , λ2 , λ3 to be distinct.

Figure 1: Reduced-form errors et

17

Electronic copy available at: [Link]

 We firstly proceed by estimating λ1 , λ2 , λ3 which then yields λˆ1 = 0.20, λˆ2 =
12.66, λˆ3 = 2.32 with standard errors of 0.04, 2.50, and 0.46, respectively.
Then based on Pairwise Wald Test, all the three lambdas are significantly
statistically different which then facilitates testing for the zero restrictions in
equation (1). We proceed by testing bˆ12 = 0, bˆ13 = 0, bˆ21 = 0, and bˆ23 = 0.
The null hypothesis of bˆ13 = 0, bˆ21 = 0, and bˆ23 = 0 cannot be rejected at any
significance levels. Meanwhile, the null hypothesis of bˆ12 = 0 is rejected at
5% level, which then supports for the assumption of tax being firstly decided
before the spending. However, assuming bˆ21 = 0andbˆ23 = 0 doesn’t alter
the resulted impulse responses. Hence, we conjecture that our model here is
statistically robust.

4.3. Data and variable construction

There are two data sets used in this paper. The first set is used to estimate
the aggregate model covering quarterly data series spanning from 1969:II to
2020:IV consisting of total government spending (central government and
regional governments), total government revenues (tax and non-tax), real
GDP, and GDP deflator. Both GDP and GDP deflator series from 1969:II
to 2000:IV are disaggregated from yearly data series using Chow and Lin
(1976)’s method using Consumer Price Index as the reference series.
The second data set is used to estimate the sub-models to go into detail
computing the impacts of specific type of fiscal to particular components
of GDP. This data set consists of quarterly fiscal series and macroeconomic
variable series spanning from 2000:I to 2020:IV. The fiscal series include
income tax, VAT, current expenditure, and capital expenditure, while the
macroeconomic series comprise real GDP, nominal GDP, real consumption,

18

Electronic copy available at: [Link]

and real gross fixed capital formation (investment). All the fiscal series are
sourced from the Indonesian Ministry of Finance, while the macroeconomic
series are sourced from the Indonesian Bureau of Statistics (BPS). When
entering the model for estimation, all variables are seasonally adjusted using
the standard X11 procedure and in logarithmic form. The fiscal series are
deflated by the GDP deflator to obtain the real terms.

5. Results and discussion: assessing the impacts of government

spending and tax

5.1. Estimating long run and short run fiscal multipliers

We divide fiscal multipliers into two, i.e., long-run and short-run multi-
pliers. The long-run multiplier results from dividing the whole cumulative
impulse responses of GDP to a fiscal shock by cumulative impulse responses
of fiscal variables, while the short-run multiplier is during the first four quar-
ters only. The long-run fiscal multiplier is presented in Panel A of Table 2
and the short-run one is shown in Panel B of the same table.
We find a significantly positive fiscal multiplier from a government spend-
ing shock and a considerably negative fiscal multiplier from a tax shock.
Comparing our results with existing literature, our results here are more de-
sirable than, for instance, in Tang et al. (2013) with a positive multiplier
from a tax shock and a negative multiplier from a spending shock, despite
the latter one is not statistically significant. In terms of the size of the mul-
tipliers, our results here are more supportive for the Neoclassical model as
the multiplier magnitude is below 1.

19

Electronic copy available at: [Link]

 Panel A. Long run fiscal multipliers (m)1
Description Lower bound m Upper bound
Multiplier of g 0.0198 0.0220 0.1187
Multiplier of t -0.2929 -0.0440 -0.0088
Panel B. Short run fiscal multipliers
Description 1 quarter 2 quarters 3 quarters 4 quarters
Multiplier of g 0.00019 0.00037 0.000720 0.00110
Multiplier of t -0.00012 -0.00204 -0.00347 -0.00451

Table 2: Fiscal multipliers of g & t to GDP

As far as the magnitudes in Panel B and Panel A of Table 2 are con-

cerned, we see a delay for fiscal having effects on the economy. We can also
see an asymmetrical effect between tax and spending shocks. If we refer to
the largest effects in Panel A of Table 2, the multiplier of tax to GDP is ap-
proximately -0.29, while the multiplier of spending to GDP is approximately
0.12, even less than half of the former. Hence we can conjecture that along
the period of sample study (1969:II to 2020:IV), the tax tool is more powerful
than the spending tool.

5.2. Effects of government spending (g) and tax (t) shocks to GDP

In this subsection, the dynamic effects of fiscal shocks on GDP are ana-
lyzed by assessing the impulse responses. They are presented together with
68% standard Bootstrap confidence intervals of Kilian and Lütkepohl (2017)
with 2000 replications. In particular, we compare the effects of one stan-
dard deviation of spending shock and one standard deviation of negative tax
shock. From Figure 2, 2.79% reduction in tax is followed by a significant

20

Electronic copy available at: [Link]

increase in GDP by approximately 0.0043% peaking up just after 1 quarter.
Meanwhile, 6.85% increase in spending is followed by an increase in GDP
by around 0.0038% and significant only after 2 quarters peaking up at the
fifth quarter before fading away afterward. One takeaway from this finding is
that a tax shock is transmitted to the economy much faster than a spending
shock. If we consider the magnitudes of both tax and spending shocks, it
also sheds that the role of spending is dwarfed by the role of the tax.

Figure 2: Impulse responses of real GDP to spending and tax shocks

21

Electronic copy available at: [Link]

5.3. Historical contribution of g and t

A historical contribution of the government spending and tax is analyzed

by extracting information from the moving average representation in the
form of historical decomposition derived from equation (1). As before, we
here focus on government spending and tax. Figure 3 plots the series of
linearly detrended real GDP in logarithms alongside the extracted structural
innovations of government spending and tax. By taking out the trend, we
can then focus on analyzing the fluctuations.

Figure 3: Historical decomposition of real GDP with respect to g and t

Spanning from 1970 to 2020, we can break down into several major

22

Electronic copy available at: [Link]

regimes, i.e. (i) before 1983, (ii) between 1983 and AFC 1998, (iii) be-
tween AFC 1998 and GFC 2009, (iv) between GFC 2009 and Covid 19, and
(v) during the Covid 19. Before 1983, the period of the first tax reform,
government spending seemed to have nothing to do with the economic per-
formance, while the tax was a distortion to the economy. During the period,
the tax system was an official assessment system where all tax matters were
calculated, imposed, and handled by government officials. The year of 1983
marked a dramatic change in the Indonesian tax system. The official assess-
ment system was abolished and replaced with a self-assessment system where
taxpayers calculated, paid, and reported their taxes by themselves.
The new self-assessment system had brought a significant impact to the
economy as indicated by the positive contribution of tax shock between 1993
and AFC 1998 (See Figure 3). Meanwhile, the slump of the government
spending shock in 1985 might be mostly associated with the highly unsettled
and tense political situation when the New Order celebrated their two-decade
in power (See Ghoshal (1986) for details). Other than the 1985 period, both
government spending and tax positively contributed to the economy until
the AFC 1998 hit.
Between AFC 1998 and GFC 2009, the contribution of tax shock was
aligned to the economic fluctuations, while the contribution of government
spending shock was nearly muted. One possible explanation from standard
macroeconomic literature is the regime of exchange rates. Literature suggests
that a fiscal policy is more effective when the regime of exchange rates is
the fixed-exchange-rates regime. The picture in Figure 3 is consistent with
this view where we see a positive contribution of the government spending

23

Electronic copy available at: [Link]

shock during the periods of the fixed-exchange-rates regime before AFC 1998.
Transmitting the positive impact of government spending on the economy in
aggregate becomes more challenging under a floating-exchange-rates regime.
Another thing to highlight during the period between AFC 1998 and GFC
2009 from Figure 2 is that the output fluctuation was partly associated with
distortion from tax.
The period after the GFC 2009 is signified by a stronger role of tax, while
the contribution of government spending remains subtle. As described in
the previous section, there were substantial changes in the tax laws during
this period such as the income tax law and VAT law were some exemptions.
Together with the stimulus package during the GFC 2009 displayed in Table
1, they might explain the positive contribution of tax during this period. In
terms of the subtle contribution of the government spending, we argue that
due to the gasoline subsidies which were just removed in 2014, mandatory
spending, and interest payments. Alongside the decentralization program
and some other earmarked spending, they might not create a positively sig-
nificant multiplier effect on the whole economy.
Now, how about the contribution of fiscal to the economy in aggregate
during the Covid 19 crisis? Note that, our analysis here is based on the
extracted structural shocks from the aggregate government spending and
taxes, not only based on the Rp695.2 trillion of the PEN program in 2020.
We spotlight Figure 3 focusing on the period 2019-2020 only. We sum up
contributions from the government shock and tax shock so that we obtain
the total contribution of fiscal to the economy, represented by the black dash
line in Figure 4.

24

Electronic copy available at: [Link]

 Figure 4: Historical decomposition of real GDP 2019-2020 with respect to g and t

There are two things that we can infer from Figure 4, i.e. first, tax
tool is more impactful than spending tool and second, it takes time for the
government disbursement to have an impact on the economy. The more
impactful tax tool by no means makes sense given the fact that some portion
of government spending has been earmarked for specific programs. Figure 4
also provides hints that tax fluctuations mostly coincide with the economic
fluctuations which are partly due to the consequence of the self-assessment
system. The trending up contribution of the government spending (g) in
Figure 4 during the second half of 2020 was consistent with Temenggung

25

Electronic copy available at: [Link]

et al. (2021) who pointed out that most of the PEN programs were realized
between July and December 2020.

5.4. Policy counterfactual analysis: How would be the GDP fluctuations

without g and t?

There are always judgments of any actions of economic agents. Then one
might raise a question how could we evaluate the actions by a fiscal authority?
Usually, there are biases when coming to judgments. The government as the
social planner in the economy, tends to defend its actions, while the opponent
will always criticize no matter what the government has done. Deriving the
moving average representation from the structural parameters in equation
(4) makes it possible to evaluate how the fiscal authority performs. Here,
they are presented in the form of historical decompositions in Figures 5 and
6.
Referring to both figures, there are good news and bad news. The good
news is that there were times when the fiscal authority performed well. In
the figures, it is represented by the fact that the GDP would have been
lower without a government spending shock or a fiscal shock. It can be
seen, for instance, between 1990 and 1998 in Figure 5 where the GDP would
have been lower without the government spending shock, owing to the fixed-
exchange-rate regimes too. Nonetheless, we had seen the limitations of the
fixed-exchange-rates regime in the AFC 1998. Another example of good fiscal
performance was between 2010 and 2018 in Figure 6 where GDP would have
been lower without a tax shock.
After the good news, now we come to the bad news and it is that the fiscal
authority did not perform well or not good enough to lift up the economy.

26

Electronic copy available at: [Link]

 Figure 5: Historical counterfactual decomposition of GDP fluctuations without g

One example is the period of 2000-2010 in Figure 6 where the GDP would
have been higher without tax shock. Another example is the period between
2010 and 2018 in Figure 5 where GDP would have been higher without a
government spending shock. However, in the latter case, the unimpressive
performance in spending had been compensated by the performance in tax.

27

Electronic copy available at: [Link]

 Figure 6: Historical counterfactual decomposition of GDP fluctuations without t

6. Policy implication

6.1. General recommendation

Before 1998, government spending seems to perform well in promoting

higher economic growth. It can be explained by the large size of spend-
ing relative to the economy and the big portion of capital spending in the
government budget. However, the declining trend contribution of non-tax
revenue and relatively stagnant tax revenue cause the lower size of govern-
ment spending. This can narrow down fiscal space which undermines the role
of fiscal role to influence the economy. However, tax revenue in Indonesia

28

Electronic copy available at: [Link]

might contribute to promote growth especially after 2014 because of the rel-
atively generous tax policy in Indonesia. It is estimated that tax expenditure
in Indonesia is more than 1,5% of GDP on average in the period 2016-2019
(BKF 2020).
Based on this experience, two main fiscal policy implications might be im-
portant to be considered by the fiscal authority. First, government spending
might not be the ultimate tool to boost economic growth, especially with the
modest size of the government spending on the economy. However, a higher
portion of capital spending most likely will perform well to promote higher
growth. Second, since tax revenue has a negative multiplier in the economy,
less offensive tax collection might improve private sector development. How-
ever, this condition will not last long so the government should improve the
tax collection in line with the economic cycles, especially to increase the fiscal
space that is available to finance capital spending.

6.2. Policy counterfactual simulation: reducing current spending vs raising

VAT

Here, we simulate the effect of reducing current expenditure versus raising

VAT rate from 10% to 12%. We measure the size of the shock as 20 percent-
age point and re-estimate equation (4) by specifying zt in (1) consisting of
current expenditure, VAT, and consumption. Using a similar procedure, the
estimated structural restrictions are as follows:
   
bˆ11 bˆ12 bˆ13 0.2080 0.0505 0
   
ˆ
−1 ˆ ˆ ˆ
B0 = b21 b22 b23  =  0
  
(6)
0.2 1.5 
   
bˆ31 bˆ32 bˆ33 0.0017 −0.0084 0.0024

29

Electronic copy available at: [Link]

where bˆ22 = 0.20 is the size of the VAT shock and bˆ23 = 1.50 is a measure for
mean of elasticity of VAT to consumption after excluding some outliers. By
setting bˆ21 = 0 and estimating bˆ12 = 0.0505, it means that we assume that
current spending reacts contemporaneously to a VAT shock, but not the other
way. We then proceed by analyzing the impulse responses of consumption to
a reduction in current spending and an increase in VAT.

Figure 7: Impulse responses of consumption to lowering current expenditure vs raising

VAT with 68% Standard Bootstrap confidence intervals of Kilian and Lütkepohl (2017)
with 2000 replications

Consulting Figure 7, it seems that consumption reacts strongly to a VAT

shock. While VAT increase does badly to consumption, cutting current ex-

30

Electronic copy available at: [Link]

penditure may not hurt it as hard as the VAT does. This implies that raising
VAT tariffs might not be a good policy option, especially when the gov-
ernment wants to maintain aggregate demand in the period of a pandemic.
On the other hand, to reduce fiscal pressure, lowering government current
expenditure might create a better result.

6.3. Policy counterfactual simulation: lowering income tax vs raising capital

expenditure for investment

Here, we simulate the effects of boosting capital expenditure versus low-

ering income tax for investment. We define capital expenditure here as the
total capital expenditure of the central government and a portion of regional
transfers and village funds allocated for capital expenditure. We then aug-
ment a variable of oil price into equation (1) as an exogenous variable which
then makes it enter each equation, while no feedback from the endogenous
variables to the oil price. For the oil price series, we use the crude oil price
index of a simple average of three spot prices (Dated Brent, West Texas
Intermediate, and the Dubai Fateh). It enters the model in logs after be-
ing seasonally adjusted. We re-estimate equation (4) by specifying zt in (1)
consisting of capital expenditure, income tax, and gross fixed capital for-
mation (investment). Using the similar procedure, the estimated structural
restrictions are as follows:
   
bˆ bˆ bˆ 0.1636 0.0468 0
 11 12 13   
Bˆ0−1 = bˆ21 bˆ22 bˆ23  =  0 (7)
   
1.4540 1.48 
   
ˆ ˆ
b31 b32 b33ˆ 0.0038 −0.0140 0.0172

31

Electronic copy available at: [Link]

where bˆ23 = 1.48 is a measure for mean of elasticity of income tax to in-
vestment after excluding some outliers. By setting bˆ21 = 0 and estimating
bˆ12 = 0.0468, it means that we assume that capital expenditure reacts con-
temporaneously to an income tax shock, but not the opposite. We then
proceed by analyzing the impulse responses of investment to an increase in
capital expenditure and a decrease in income tax.

Figure 8: Impulse responses of investment to boosting capital expenditure vs lowering

income tax with 68% Standard Bootstrap confidence intervals of Kilian and Lütkepohl
(2017) with 2000 replications

Referring to Figure 8, it seems that increasing capital expenditure will

have a more persistent positive effect on investment. The effect dissipates

32

Electronic copy available at: [Link]

just after 40 quarters or 10 years. Lowering income tax also has a significantly
positive effect on investment with even a larger effect than increasing capital
expenditure. However, the effects fade away just after 4 quarters or one year.
To be more comparable in terms of their size effects, we proceed by com-
puting the multipliers of both shocks on investment as presented in Table 3.
They are based on cumulative impulse responses of 40 quarters or 10 years.
The impulse responses in Figure 8 together with figures in Table 3 highlight
that the capital expenditure effect is long-lasting, while the income tax effect
matters in short term only. In addition, we label the third shock in Figure
8 as other shock since we do not have sufficient information to label it as a
specific shock. It is merely any shock other than fiscal shocks affecting the
investment.

7. Conclusions

We find an overall positive fiscal multiplier of government spending and

a negative fiscal multiplier of tax as predicted by theory. With their mag-
nitudes of below 1, they are more supportive of the Neoclassical than the
Keynesian view. We also find that tax has a larger multiplier than spending.
Moreover, through a policy counterfactual study, we simulate the effects of
reducing current expenditure versus raising VAT on consumption and raising
capital expenditure versus lowering income tax on investment. The former
reveals that VAT increase has more adverse effects on consumption than a
reduction in current expenditure. And the latter predicts that lowering in-
come tax has a larger effect in short term, but raising capital expenditure
promotes longer-lasting effects than lowering income tax which aggregates

33

Electronic copy available at: [Link]

larger effects for the economy in the long run.
Given the macroeconomic framework used in this paper, there are limi-
tations that should be kept in mind with care. A number of issues that may
influence the performance of fiscal policy and have not been covered here are
for example the timing for the policy announcement, the credibility of the
fiscal authority, and other institutional factors. Hence, we leave those for
further research.

References

Abdurohman (2013). Countercyclical Fiscal Policy in Indonesia. PhD thesis,

Australian National University.

Alesina, A. and Ardagna, S. (1998). Tales of fiscal adjustment. Economic

Policy, 13(7):488–545.

Bank, W. (2020). Indonesia Public Expenditure Review: Spending for Better

Results. Technical report, World Bank.

Basri, M. C. (2013). A Tale of Two Crises : Indonesia ’ s Political Economy.

Batini, N., Eyraud, L., and Weber, A. (2014). A simple method to compute
fiscal multipliers. IMF Working Paper, 14(93):Washington DC: Interna-
tional Monetary Fund.

BKF (2020). Tax Expenditure Report 2020. Technical report, Ministry of

Finance of the Republic of Indonesia, Jakarta.

34

Electronic copy available at: [Link]

Blanchard, O. and Perotti, R. (2002). An empirical characterization of the
dynamic effects of changes in government spending and taxes on output.
Quarterly Journal of Economics, 117(4):1329–1368.

Canzoneri, M., Collard, F., Dellas, H., and Diba, B. (2016). Fiscal Multipliers
in Recessions. Economic Journal, 126(590):75–108.

Chow, G. C. and Lin, A.-L. (1976). Best linear unbiased estimation of missing
observations in an economic time series. Journal of the American Statistical
Association, 71(355):719–721.

Eka, I. W. A. (2019). The Impact of The Indonesian Tax Administration

Reform on Tax Compliance and Tax Revenue. Jurnal Ekonomi dan Pem-
bangunan, 27(1):1–24.

Fukuchi, A. (2020). Impact of COVID-19 on the Indonesian Economy and

Its Prospects.

Ghoshal, B. (1986). Indonesia in 1985 : A Year of Trials. Southeast Asian

Affairs, pages 99–121.

Habir, M. T. and Wardana, W. (2020). COVID-19 ’ s Impact on Indonesia’s

Economy and Financial Markets. Perspective, 142:1–13.

Hebous, S. (2011). The effects of discretionary fiscal policy on macroeconomic

aggregates: A reappraisal. Journal of Economic Surveys, 25(4):674–707.

Hemming, R., Kell, M., and Mahfouz, S. (2002). The effectiveness of fiscal
policy in stimulating economic activity - A review of the literature. IMF
Working Paper, 02(208).

35

Electronic copy available at: [Link]

Huidrom, R., Ayhan Kose, M., Matsuoka, H., and Ohnsorge, F. L. (2020).
How important are spillovers from major emerging markets? International
Finance, 23(1):47–63.

Kilian, L. and Lütkepohl, H. (2017). Structural Vector Autoregressive Anal-

ysis. Themes in Modern Econometrics. Cambridge University Press.

Lanne, M. and Lütkepohl, H. (2008). Identifying monetary policy shocks via

changes in volatility. Journal of Money, Credit and Banking, 40(6):1131–
1149.

Lee, S., Liao, Y., Seo, M. H., and Shin, Y. (2020). Desperate Times Call
for Desperate Measures: Government Spending Multipliers in Hard Times.
Economic Inquiry, 58(4):1949–1957.

Ministry of Finance of the Republic of Indonesia (2005). 2005 State Budget.

Technical report, Ministry of Finance of the Republic of Indonesia, Jakarta.

Morris, R., Ongena, H., and Schuknecht, L. (2006). The Reform and Imple-
mentation of the Stability and Growth Pack.

Murphy, K. M., Shleifer, A., and Vishny, R. W. (1989). Industrialization and

the Big Push. Journal of Political Economy, 97(5):1003–1026.

Musgrave, R. A. (1959). The theory of public finance. McGraw-Hill, New

York.

Peren Arin, K., Koray, F., and Spagnolo, N. (2015). Fiscal multipliers in
good times and bad times. Journal of Macroeconomics, 44:303–311.

36

Electronic copy available at: [Link]

Rigobon, R. (2003). Identification through heteroskedasticity. The Review
of Economics and Statistics, 85(4):777–792.

Seda, F. (2009). Kebijakan Anggaran Penerimaan dan Belanja Negara

(APBN) Berimbang dan Dinamis. In Abimanyu, A. and Megantara, A.,
editors, Era Baru Kebijakan Fiskal, Pemikiran, Konsep, dan Implemen-
tasi. Kompas.

Shafuda, C. P. and De, U. K. (2020). Government expenditure on human

capital and growth in Namibia: a time series analysis. Journal of Economic
Structures, 9(1).

Tang, H. C., Liu, P., and Cheung, E. C. (2013). Changing impact of fiscal
policy on selected ASEAN countries. Journal of Asian Economics, 24:103–
116.

Tanzi, V. and Zee, H. H. (1997). Fiscal policy and long-run growth. IMF
Staff Papers, 44(2):179–209.

Temenggung, D., Saputro, A., Rinaldi, R., and Pane, D. (2021). Managing
recovery and seizing reform opportunities. Bulletin of Indonesian Eco-
nomic Studies, 57(1):1–28.

Wijayanto, B. and Vidyattama, Y. (2017). Revenue and Distributional Im-

pact Analysis of Indonesian Personal Income Tax Reform in 2008. Litera-
ture Review, 63(2):97–113.

Yien, L. C., Abdullah, H., and Azam, M. (2019). Monetarism or Keyne-

sian Discourse in Indonesia. Bulletin of Indonesian Economic Studies,
55(3):367–385.

37

Electronic copy available at: [Link]