Versión de HTML Básico
Fiscal policy and economic growth: a simulation analysis for Bolivia
Analíti a
k
4
Revista de Análisis Estadístico
Journal of Statistical Analysis
The second column describes the deep parameters of
the production functions. The depreciation rate of private
capital
δ
was obtained by calibrating private investment
and is set at 2.8 percent per year. The output factor elastic-
ities
α
in each sector were obtained by reducing the 35 sec-
tors in Bolivia’s 2006 input-output matrix to 6 sectors: agri-
culture, hydrocarbons, mining, importables, non-tradables
and infrastructure. We are particularly interested in the
following infrastructural sectors: i) energy, gas and wa-
ter, ii) transportation and storage and iii) communications.
We then use the value-added decomposition of factor pay-
ments for 1996 (the only year available) and impute these
shares for our 2006 sectoral value added. The correspond-
ing calculations are shown in table A.1 in the appendix.
The shares of infrastructure in each sector are key pa-
rameters. We have also used our input-output matrix dis-
aggregation, but here we employed the intermediate con-
sumption of infrastructure in each of the five sectors of
the model. In other words, we have calculated the value
of
φ
in each sector as the share of intermediate consump-
tion of infrastructure in agriculture, mining, hydrocarbons,
importables and non-tradables. Recall that public capi-
tal is presumed to be free for firms in the model, so it
seems strange to calibrate each sector’s share parameters
using intermediate consumption, which is an expenditure.
We address this concern by assuming that the government
subsidizes the private sector via public goods. The govern-
ment provides this public capital, but it is produced by the
private sector. The corresponding calculations are shown
in table A.2 of the appendix.
13
The third column contains the TFP parameters. These
parameters have been calibrated to match each sector’s
share of GDP as closely as possible. The auto-regressive
coefficients and shock volatilities were set to correspond
with the autocorrelation between output and the standard
deviations of the AR(1) regression residuals for each sec-
tor’s output.
14
The fourth column shows parameters for government
and fiscal variables. The parameters of the AR(1) process
for government expenditures are taken from a simple OLS
regression, while the parameter g is calibrated to match
government expenditures as a share of GDP. The depre-
ciation rate of public capital
δ
g
has been estimated by the
World Bank at about twice that of the depreciation rate of
private capital. The benchmark effectiveness parameter
θ
is estimated here using data on the so-called “loss indica-
tors”. In particular, we use the power, telecom, roads and
water loss indicators. The Bolivian loss index across all in-
frastructure types is calculated as a weighted loss and is
then compared to the weighted average in industrialized
countries. The calculations are shown in table A.3 of the
appendix. According to these calculations, Bolivia has a
level of effectiveness of 61.3, meaning that infrastructure in
Bolivia is 39 percent less effective than in developing coun-
tries.
15
The Bolivian tax system is described by the various tax
rates applied across the economy. The consumption tax
τ
c
is approximated by the 13-percentvalue added tax (VAT).
The tax on capital income
τ
k
is also levied at a rate of 13
percent and corresponds to the Complementary Regime
Value Added tax (CR-VAT). The tax on hydrocarbons
τ
xh
is 32 percent, and is known as the Direct Tax on Hydro-
carbons (DTH). Finally, the import tariff
τ
m
represents the
average tariff for all the imported products and has a value
of 10 percent.
We display the so-called exogenous prices in column
5 of Table 3. Each of these prices follows a standard law
of motion and most of their parameters are estimated us-
ing OLS regressions. We calibrated the constant terms of
the AR(1) specifications of these relative prices using the
respective index prices calculated by the Bolivian Central
Bank. Finally, we calibrate
φ
as 0.248 to make the external
debt-to-GDP ratio equal 0.3790, which is consistent with
the capital account balance in the steady state. This value
for
φ
combined with a value of
ω
equal to 1.2 gives a coun-
try risk value equal to 0.05857.
4 Results
This section reports the various simulations we carried
out using the key model parameters. These simulations
quantify the effects of fiscal policy on macroeconomic vari-
ables such as output, consumption, investment, etc. We
also distinguish between long run effects and short run
dynamics. The long run effects are determined by com-
paring the model’s steady states in the baseline scenario
with steady states in the simulated scenarios. Determining
the effects on the short run dynamics requires that we im-
pose initial conditions, solve the model (
i.e.
, find the policy
functions of the control variables and the endogenous state
variables’ laws of motion) and characterize the transition
to the new steady state.
16
4.1 Steady state comparisons
In this subsection we present changes in the long run
steady state values for: consumption of each good (
c
m
,
c
n
),
physical production in each sector (
Y
xh
,
Y
xa
,
Y
xm
,
Y
m
,
Y
n
),
13
A better specification of the production function is
f
(
z
i
,
t
,
k
i
,
t
,
k
∗
t
)
, where
x
represents private intermediate consumption.
14
These parameter values are important for changes in the speed of convergence to the steady state.
15
We use the same weights as in Rioja (2003), namely 0.40, 0.10, 0.25, 0.25 respectively for power, telecom, paved roads and water systems. The
effectiveness index
θ
for developing countries is normalized such that a value of 1 indicates highly effective infrastructure.
16
According to our specification, the policy functions of the control variables cannot be obtained analytically, so we have to resort to numerical
methods. We used the Schmitt-Grohé and Uribe (2004) second-order approximation technique. This perturbation method has proven superior to
traditional linear-quadratic approximations.
Analítika,
Revista de análisis estadístico
, 2 (2012), Vol. 4(2): 57-79
65