A DYNAMIC MODEL OF FISCAL BEHAVIOUR FOR INDONESIA.
3.4. Econometric Estimation.
3.4.2. Results and Interpretation.
The econometric estimates on a time series basis show that aid does have an
effect on fiscal behaviour, but mainly through private investment and the dynamic
linkages stemming from lagged government consumption and income44. Tables 3.4.2.1,
3.4.2.2 and 3.4.2.3 present the results for the three alternative models: Model I uses
grants and loans, Model II bilateral and multilateral aid, and Model in total aid.
Diagnostic test are reported in appendix 5. Tables 3.4.2.4, 3.4.2.5 and 3.4.2.6 report the
calculated reduced form coefficients for each model expressed as percentage changes
following a 1% shock in aid45. These tables show aid’s effect on the budget, on GDP
and on private consumption on impact and after the second period, when dynamic
linkages have started to operate (i.e. short run and medium run effects). The full long
run dynamics of the estimated model is presented in the next paragraph.
Grants and loans do not exhibit a significant direct impact on fiscal behaviour.
However, the interaction between fiscal behaviour and private investment decisions
channels an indirect effect. Reduced form coefficients show that grants tend to
discourage taxation efforts and to deflect resources away from public investment and to
44 Instrumental Variables estimations were carried out with PcGive 8.0 while 3SLS estimations were implemented with PcFiml 8.0.
45 The corresponding reduced form coefficients in levels are reported in Appendix 3.
government consumption. On the contrary, loans positively affect government revenues
and investment, although in the short run the impact on taxation is negative. However,
the correct interpretation of the results should be based on statistically significant
parameters, as pointed out in paragraph 3.2.3. The tables show both sets of results to
highlight how misleading it would be to rely on all parameters, significant and
insignificant.
In fact, if we consider significant parameters only, there is no evidence of a
short term impact either on fiscal behaviour or on income and consumption. Once
dynamic linkages initiate a multiplier process, statistically significant effects emerge.
At the end of the second period following the aid shock, a 1% increase in grants
discourages taxation efforts and leads to a 4% drop in revenues. Similarly, on the
expenditure side, investment and consumption will be curtailed, although investment
will be reduced by less than consumption. Thus, grants seem to partly finance a
tightening in fiscal policy aimed at reducing the size of government and exhibit a pro-
investment bias. As for loans, the effect is reversed. A 1% rupiah increase in loans
results in 1.7% increase in tax revenues leading to higher public investment and
consumption. Even if public consumption grows more than investment, investment still
increases more than proportionately with respect to the increase in loans. This implies
that loans pro-consumption bias is not high enough to cause fungibility.
An explanation for the opposite effects of grants and loans may lie in the
disincentive effect of receiving non-repayable grant money, which are used by policy
makers to reduce the tax burden. Given the budget constraint, reduced finances result in
reduced outlays and a general depressing effect on the whole economy. On the other
that they induce an intensification of tax collection efforts. Although increased
revenues will finance public investment less than public consumption, the whole
economy will benefit as a result of the multiplier process. As shown in table 3A.2.4 to
3.4.2.6, the effect on income and private consumption are both negative in the grant
case and positive in the loans case. The size of the coefficients in general is due to the
small figures for aid as a proportion of GDP, which ranged from 1% to 6.2%. In
particular, grants never exceeded 1% of GDP and loans ranged between 1% and 5.2%.
Interestingly, the use of the GOV dataset, yields different results. The impact of
aid is direct and not channelled via private investment. Grants have a significant
negative impact, while loans exhibit insignificant effects.
Model II and Model IE perform less well than Model I. The impact effect of aid
is only direct as coefficients for bilateral, multilateral and total aid in the private
investment equation are insignificant. Moreover, there is no evidence of long run
effects on the budget, on income and on private consumption from multilateral aid. In
both models, bilateral and total aid negatively affect, on impact, only taxation, public
consumption and income, while there is no statistical evidence for their effects on
public investment and private consumption. However, at the end of the second period,
not only T and Gc but also Ig will be curtailed as a consequence of an increase in
bilateral or total aid. In both cases, there is a pro-investment bias, given that
government investment will be reduced by .8% and .9%, respectively, against a fall in
public consumption of 1.3% and 1.2%, respectively.
Once again results from the corresponding regressions using the GOV dataset
differ. Government taxation efforts are significantly influenced by bilateral and
multilateral aid. Bilateral aid has a direct and negative impact only. On the contrary,
multilateral aid positively affects taxes, both directly and indirectly. Surprisingly, public
investment does not interact with private investment decisions and the only aid effect
comes from bilateral aid and is negative. When we turn to total aid results the only
significant effect is the direct negative impact on taxation efforts.
As for general diagnostics, the three model perform well when estimated with
the IMF-OECD dataset. The estimated equations (restricted reduced forms of the
structural model) appear to parsimoniously encompass their respective unrestricted
reduced forms. There is also no evidence of problems of autocorrelation and
misspecification. On the contrary, for all the three models estimated with GOV dataset,
diagnostic tests do not allow us to accept the null of the equations parsimoniously
encompassing the unrestricted reduced forms. There are also problems of
autocorrelation in both the private and the public investment equations when bilateral
and multilateral aid are used.
The comparison with 3SLS results in all cases tend to confirm the
corresponding results from IV estimations. However, these model perform badly in
terms of some diagnostic tests. The most serious problem is the rejection in all cases of
the null that the reduced model parsimoniously encompasses the system. There are
Finally, it is worth noting the robustness of the private consumption equation.
The long run estimated marginal propensity to consume out of disposable income is
0.62 for the IMF-OECD dataset and 0.60 for the GOV dataset46.
Table 3.4.2.I. Model I. Grants and Loans (IMF-OECD Dataset)
Constant I imf *D out,., r «e,Himf A, A,
rpimf 2661.0**
(1239.7) .37“(.07) .46”(.19) .65“(.28) -3.04(3.24) (1.07)-.75 1 imf
** -932.6
(876.5) (.05).09’ .31“(.13) .60"(.20) -2.78(2.29) 1 .0 2(76) Instruments used: Ag,,.i, A|,t.i, O Ft_i, Y u , Dcred.
Constant Oil tv, r 'Jc.l-limf A, A, A«.i-i A|,m OF,., Y vi Dcred
■ imf -10098**
(1793.6) -.03(22) -.50(.41) (3.72)6.14 (1.30).30 -14.3“(5.68) 3.67“(1.48) -.80"(27) .37“(.06) .15“(04) Private Consumption Equation: Long Run
C = 10990“ Constant +.62“ Yd‘mf
(2460) (.05)
*, ** denote significance at 10% and 5% respectively. Standard errors in parentheses.
46 Woo, Glassburner and Nasution (1994) obtain a 0.63 elasticity of the private consumption deflator with respect to the GDP deflator. This is loosely comparable with our estimated marginal propensities to consume, 0.62.
Table 3.4.2.2. Model II. Bilateral and Multilateral Aid (IMF-OECD Dataset)
Constant I « o u t,., P «C.Uimf Ab Am
rjiimf 2893.4" .41" .47" .47' -1.63" 3.47
(1200.1) (.07) (.18) (.26) (.74) (315)
I ,“"' -634.2 .08 .34" .49" .26 .005
(903.2) (.05) (.14) (.20) (.56) (2.37)
Instruments used: Ab,,_i, OF,.,, Y,.i, Dcred.
Constant Oilt,., g~* t-1imf Ab Am Ab.t-i Ani,|.i O F,., Y ,., Dcred
■ imf
*P -6858.3**(1915.4) -.0 2
(.25) (61).23 -1.16(1.44) (6.93)3.99 (1.65)2.49 -9.78(7.87) -.52(32) .23"(.08) .17"(05) Private Consumption Equation: Long Run
C = 11040* Constant +.62** Yd1"*
(2581) (.05)
*, ** denote significance at 10% and 5% respectively. Standard errors in parentheses.
Table 3.4.2.3. Model III. Total Aid (IMF-OECD Dataset)
Constant i.“ o u t,., G«,,.,“ A
rjiiltlf 2989.3" .38" .47" .56" -1.28'
(1214.7) (.07) (18) (.26) (72)
i f * -719.5 .08 .34" .50" .28
(881.5) (.05) (13) (.19) (53)
Instruments used: A,.,, OF,., Y u , Dcred.
Constant out,., r imf A A,., O F,., Y i-i Dcred
« imf
*P -8178.7**
(1693.3) -.04(25) -.33(.44) (1.19).0 2 (1.25)1 .2 1 -.69"(.30) .31"(.06) .14"(.05) Private Consumption Equation: Long Run
c = 10980** Constant +.62** Yd“"'
(2449) (.05)
Percentage Changes Following a 1% Increase in Aid.
Table 3.4.2.4. Model I. Grants and Loans (IMF-OECD Dataset (.Estimated Impacts on the Budget, on GDP and on Consumption.
All Parameters 1“ period 2nd Period Significant Parameters 1" Period 2nd Period a T / d A , A I TK R -.7 -3.1 - -4.7 -5.0 -4.0 - -2.9 d Gc / d A g Ag /g c 3.3 -.9 - -5.4 a T I d A, A, I T -.8 .3 - 1.7 d l g I d A, A, l l g 3.3 3.0 - 1 .1 a Gc I d A, A, / Gc -1 . 2 .5 - 1.8 d Y / d A g Ag / Y 1 .0 -4.5 - -5.7 d Y I d A, A, / Y -.04 1.5 - 1.7 d C / d A g Ag 1C .4 -3.8 - -4.7 d C I d A, A, 1C .0 2 1 .2 - 1 .2
Table 3.4.2.5. Model II. Bilateral and Multilateral Aid (IMF-OECD Dataset). Estimated Impacts on the Budget, on GDP and on Consumption.
Percentage Changes Following a 1% Increase in Aid. All Parameters
1“ period 2nd Period
Significant Param eters
1” Period 2nd Period 3 T l d A h A J T -2.5 -2.4 -1.9 -2.4 d l g / d A h A j l g .5 -1 .0 - -.9 d Gc 1 d Ah Ah / Gc -2 . 1 -1 . 1 -1.0 -1.3 a T / d A m Am / T 3.6 3.3 - - d l , / d A m An I I g .7 5.4 - - 3Cc l d A m Am / G c 8 .1 4.5 - - a Y I d Ah A J Y -.5 .4 - . 2 -.4 d Y l d A m Am / Y 1.5 -2.7 - - d C / d A h Ah I C - . 1 .6 - - . 1 d C / d A m Am / C .2 -3.2 - - 127
Percentage Changes Following a 1% Increase in Aid.
Table 3.4.2.6. Model III. Total Aid (IMF-OECD Dataset). Estimated Impacts on the Budget, on GDP and on Consumption. Elasticities.
All Parameters 1“ period 2nd Period Significant Parameters 1" Period 2nd Period 3 T / d A A I T -1.4 -1.4 -1.4 -1 .8 a l t I d A A l l g .8 .2 - -.8 a Gc I d A ■ A I Gc -.9 -.5 -.4 -.7 a Y I d A A l Y -. 2 .1 - . 2 -.6 3 C / d A A I C .0 2 .3 - -.3 3.4.3. A Simulation Exercise.
The estimated parameters are also used to simulate the impact of grants and
loans on fiscal and macroeconomic variables in order to explore the full long run
dynamics of the estimated model. Graphs 3.2.3.1 to 3.4.3.6 show the simulated effects
that temporary and permanent shocks in grants, loans and joint grants and loans have
on all the endogenous variables of the system. We carried out simulations47 using
shocks of the size of the estimated standard errors from autoregressive regressions of
grants and loans, namely 69.32 billion rupiahs for grants and 279.27 billion rupiahs for
loans. Percentage changes are plotted against time.
Graph 3.4.3.I. Effect of a Temporary Shock in Grants.
Y;--- I p : ---- c:
Graph 3.4.3.2. Effect of a Temporary Shock in Loans.
Y:--- Ip:--- C:
% Change on the vertical axis; time on the horizontal axis.
Graph 3.4.3.3. Effect of a Temporary Joint Shock in Grants and Loans.
Y:--- Ip :--- C:
% Change on the vertical axis; time on the horizontal axis. Graph 3.4.3.4. Effect of a Permanent Shock in Grants.
Y:---- Ip:---- C:
Graph 3.4.3.S. Effect of a Permanent Shock in Loans.
Y :--- I p — C:
% Change on the vertical axis; time on the horizontal axis.
Graph 3.4.3.6. Effect of a Permanent Joint Shock in Grants and Loans.
Y;--- Ip :— C:
% Change on the vertical axis; time on the horizontal axis.
Simulations confirm the preceding discussion of the results. We note briefly the
overall negative impact of grant aid in contrast to the generally positive impact of loans.
Their joint impact is also positive, except for a 2 periods adjustment in taxation efforts.
What is striking is the size of the short period impact in all cases. This may be ascribed
once again to the relative little weight of grants and loans to GDP. In fact, the
temporary effects tend to stabilise to values not in excess of 2.5%. Not surprisingly,
private and public investment are the most sensitive to such shocks, given the structure
of the model. On the budget side, the policy maker in this simulated economy would
reduce taxation efforts for about one year as a response to increased loans and joint
loans and grants. Afterwards, tax revenues would rise again. The increase in public
investment would therefore be at the expenses of taxation efforts only in the very short
run. Grants alone would only harm the budget and the overall economy. Similar
behaviour can be inferred when considering the reactions to permanent shocks. Not
surprisingly, these cause a much higher impact on the economy, especially on
investment decisions, both private and public.