Applied Economics Letters, 2003, 10, 299–302

Optimal monetary rules: the case of Brazil

C H A R L E S L I M A D E A LM E I D A
, M A R C O A U R É L I O PE R E S # ,
G E R A L D O D A S I L V A E S O U Z A $ an d B E N J A M I N
MIRANDA TABAKô

Universidade de Brası´lia, #Secretaria de Orçamento Federal, $Departamento de
Estatı´stica, Universidade de Brası´lia and ôDepartamento de Estudos e Pesquisas,
Banco Central do Brasil, SBS, Quadra 3, Bloco B, Ed. Sede, 9 andar, 70074-900,
Brası´lia

Within a dynamic programming approach, an optimal rule for the central bank to
attain its inflation targeting goals is derived. The short-run nominal interest rate is
used as an instrument to achieve monetary objectives. The model is tested for the
Brazilian economy and compared with results found for other countries. Evidence
for the estimated feedback interest rule for the Central Bank suggests that the cost of
reducing inflation in an open economy is lower than that of a closed economy.

I. INTRODUCTION and to estimate a short-run reaction function for interest

rates for Brazil, for an open economy. The main
Recently, several countries have been adopting a target assumption is that interest rates are the central bank’s
inflation framework for monetary policy. New Zealand, main instrument to reduce inflation and the level of
Canada and the UK have decided to employ an inflation activity.
targeting framework in the conduit of monetary policy and Normally, models estimating IS-AS-type models use
have successfully reduced their inflation rates and gained OLS regressions. In this paper we derive an optimal
control on inflation.1 It is often argued that an independent feedback rule and use the estimated coefficients from IS-
central bank is a key element for a successful monetary AS equations to find an empirical relation between central
policy. The good performance in the maintenance of low bank’s instrument and macroeconomic variables such as
inflation rates that the National Bank of Switzerland and inflation, output gap and exchange rates. The use of two-
the Bundesbank have had is mainly attributed to their high stage least squares using adequate instrumental variables to
level of independence, and it has certainly strongly estimate these equations is proposed to overcome problems
influenced the position of the system of European central inherent to the nature of Brazilian macroeconomic vari-
banks toward an independent central bank. ables.2 Results obtained here are compared with those
Following this trend, other countries (e.g. Chile, Mexico, found in the literature.
Argentina, Spain and France) have been granting a greater The plan of the paper is as follows. In the first section a
independence to their central banks allowing them to Taylor rule is derived for the Brazilian economy. Section II
conduct monetary policy with much less government inter- presents empirical results. The final section provides
ference. The main goal is price stability and a common conclusions and gives directions for further research.
point among these countries has been the adoption of an
inflation targeting framework. In Brazil, since mid-1999, six
months after abandoning the fixed exchange rate regime, II. THE MODEL
the Central Bank of Brazil has adopted an inflation
targeting regime for the conduit of monetary policy. In this section an optimal monetary rule is derived for the
The main purpose of this paper is to derive an optimal Brazilian economy using the following IS equation:
feedback rule based on the model proposed by Ball (1998) ytþ1 ¼ a1 yt  a2 it þ a3 et þ utþ1 ð2:1Þ
1
For an interesting analysis of inflation targeting see Walsh (2001).
2
Using instrumental variables for the terms containing lagged inflation is crucial because these terms are correlated with the residuals.

Applied Economics Letters ISSN 1350–4851 print/ISSN 1466–4291 online ß 2003 Taylor & Francis Ltd 299
http://www.tandf.co.uk/journals
DOI: 10.1080/0003684032000066804
300 Charles Lima De Almeida et al.
where yt denotes the output gap, it is the real interest rate, Solving problem (2.11) with respect to t gives the first-
et is the real exchange rate, and ut is a demand shock, order condition:
assumed to be normally distributed.
The supply curve is represented by the traditional t þ Vz Et ðztþ1 Þ ¼ 0 ð2:12Þ
Phillips curve:
Applying the envelope theorem with respect to zt gives:
tþ1 ¼ t þ yt þ ð"t  "t1 Þ þ tþ1 ð2:2Þ
Vz ðzt Þ ¼ zt þ Vz Et ðztþ1 Þ ð2:13Þ
where t is the inflation rate,
"t is the depreciation rate in
Multiplying Equation 2.13 by , substituting in Equation
the nominal exchange rate and tþ1 the supply shock not
2.12, taking this expression one-step forward and
correlated with utþ1.
expectations:
The policy maker chooses in instant t the interest rate it,
and the state variable in instant t is
Et Vz ðztþ1 Þ ¼ zt þ t  Et ðtþ1 Þ ð2:14Þ

zt ¼ yt þ t þ ð"t  "t1 Þ ð2:3Þ
Inserting Equation 2.14 in Equation 2.12:
The optimal feedback rule will be given by
 
t ¼ Xzt ð2:4Þ t ¼  zt þ Et ðtþ1 Þ ð2:15Þ
þ 2 þ  2
where When the policy is established in instant t, zt is the state
variable and thus the optimal policy rule has a quadratic
t ¼ a1 ytt þ a2 it þ et ð2:5Þ form t ¼ Xzt. Therefore

Equations 2.1 and 2.2 can be rewritten as Etþ1 ðtþ1 Þ ¼ XEt ðztþ1 Þ ¼ X ð1 þ X Þzt ð2:16Þ

ytþ1 ¼ t þ utþ1 ð2:6Þ Inserting this expression in Equation 2.15 gives the
following quadratic form:
and 
X 2    þ  2  X þ  ¼ 0 ð2:17Þ
tþ1 ¼ zt þ tþ1 ð2:7Þ

Stability requires that

1 þ 2ð1 þ
ÞX
< 1. Hence, the
We assume that the central bank’s loss function is given by solution to Equation 2.17 is given by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 X 1  
ð   þ  2 Þ  ð   þ  2 Þ2 þ 4ð 2 2 Þ
L ¼ Et i y2tþ1 þ 2tþ1 ð2:8Þ
2 i¼1 X¼
2
ð2:18Þ
The objective of the policy maker is to minimize this loss
function subject to remembering that
ztþ1 ¼ zt þ t þ tþ1 þ utþ1 ð2:9Þ ztþ1 ¼ zt þ t ¼ ðX þ 1Þzt ð2:19Þ

Define the value function as After some algebraic operations the product of the roots is
 
1 1
Vðzt Þ ¼ minEt ð y2tþ1 þ 2tþ1 Þ þ Vðztþ1 Þ ð2:10Þ X1 X2 ¼  < 0 ð2:20Þ
2

replacing Equations 2.6, 2.7 and 2.9 in the value function The root of interest is the one that satisfies the stability
we obtain: condition, that is the negative root X2. Finally, replacing
X2 in Equation 2.4 gives

1 1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Vðzt Þ ¼ min Et ðt þ utþ1 Þ2 þ Et ðzt þ tþ1 Þ2
t 2 2 ð   þ  2 Þ  ð   þ  2 Þ2 þ 4ð 2 2 Þ
 ð2:11Þ t ¼ zt
2
þEt Vðzt þ t þ utþ1 þ tþ1 Þ
ð2:21Þ
Optimal monetary rules: the case of Brazil 301
We can derive the optimal rule for the interest rate Applying Augmented Dickey and Fuller tests, the null of
a unit root for the output gap, interest rates and inflation is
a1 X2 a3 X2 X2 rejected. Results for these unit roots are available upon
it ¼ yt þ
et þ t þ  et ð2:22Þ
a2 a2 a2 a2 request from the authors.
In order to derive the optimal policy rule for the
Brazilian economy,  ¼ 0.7 and ¼ 1 were assumed, which
are the intertemporal discount factor and the relative
III. EMPIRICAL RESULTS weight of output gap in the loss function. After replacing
these parameters and coefficients in Table 1 and Table 2
For the econometric analysis we have used quarterly data one obtains
and the sample begins in the first quarter of 1994 and ends
in the last quarter of 2001. All variables are in natural logs. it ¼ 5; 5yt þ 4; 2t ð3:1Þ
As a proxy for the output gap we have estimated a
Hodrick–Prescott filter and used the difference between These results are quite different from those found in Taylor
observed GDP and the filtered series. The inflation rate is (1993) and Ball (1998) and are more in line with those
given by IPCA. The interest rate is given by SELIC, which found in Walsh (1997) (see Table 3).
is the instrument that the central bank uses to achieve it’s The coefficient on the output gap is similar to that found
price stability goals. in Ball (1998), while the coefficient on inflation is much
According to the results, shown in Table 1, both the lag higher than that found previously, which suggests that the
of the output gap and lagged interest rate are significant central bank of Brazil has to increase its interest rates much
in explaining current output gap, and the sign of the more than developed countries in order to counterbalance
coefficients are in line with the expected sign. As an increase in inflation.
instruments, a dummy for the Russian crisis, three lags Table 4 and Table 5 present results for an open economy.
for the interest rate and four lags for government spending Replacing the results one obtains the optimal rule:
were used.
Table 2 presents empirical results for the Phillips it ¼ 5:2yt þ 0:3t þ 0:6"t1 þ 0:2
et ð3:2Þ
equation. Both lagged output gap and inflation are
significant in explaining current inflation. As instruments, To the best of our knowledge, most research on developed
a dummy for the Russian crisis, six lags for the inflation countries has estimated different optimal feedback rules,
rate and two lags for government spending have been used. making comparisons more difficult. As can be seen, the
coefficient on inflation has decreased to 0.3. Thus, the
Table 1. IS equation – closed economy

Variables Coefficients p-value Table 3. Comparison of optimal rules – closed economy

yt1 0.34 0.04 Coefficients yt t yt1
(0.1697)
it 0.06
0.00 Taylor (1993) 0.50 1.50 –
(0.0117) Walsh (1997) 4.37 1.26 1.59
Adjusted R2 ¼ 77% Authors 5.50 4.20 –
Ball (1998) 0.80 1.46 –
Standard errors are given in parentheses.
Rejection of the
null with 99% confidence.

Rejection of the null with 95%
confidence.
Table 4. IS equation – open economy

Table 2. Phillips equation – closed economy Variables Coefficients p-value

Variables Coefficients p-value yt1 0.36

0.04

(0.168)
yt1 0.34

0.029 et2 0.04

0.05
(0.014) (0.018)
t1 0.60
0.000 it 0.06
0.00
(0.023) (0.014)
Adjusted R2 ¼ 90% Adjusted R2 ¼ 75%

Standard errors are given in parentheses.
Rejection of the null Standard errors are given in parentheses.
Rejection of the null
with 99% confidence.

Rejection of the null with 95% with 99% confidence.

Rejection of the null with 95%
confidence. confidence.
302 Charles Lima De Almeida et al.

Table 5. Phillips equation – open economy derived using a dynamic programming approach and a
dynamic loss function. IS-AS equations have been esti-
Variables Coefficients p-value mated using two-stage least squares and an optimal
yt1 0.08

0.02 feedback rule has been fitted for short run interest rates
(0.05) for both closed and open economies.

"t 0.07

0.03 It has been found that the feedback rule behaves
(0.01) differently from similar rules estimated for developed
t1 0.65
0.00 countries. For the open economy evidence suggests that
(0.038)
interest rates needs to rise less than one-to-one with inflation
Adjusted R2 ¼ 64%
(while the contrary happens within a closed economy).
Standard errors are given in parentheses.
Rejection of the null Thus, it is found that within an open economy the central
with 99% confidence.

Rejection of the null with 95% bank has much more power to reduce inflation than within a
confidence. closed economy. This issue will be left for further research.

nominal interest rate is increased more than five-to-one

with increases in output gap. The cost to reduce inflation
seems to be lower in an open economy, which is an REFERENCES
argument in favour of commercial liberalization.
Ball, L. (1998) Policy Rules for Open Economies. NBER Working Paper
5952, October.
Taylor, J. B. (1993) Discretion versus policy rules in practice, Carnegie-
IV. CONCLUSIONS Rochester Conferences on Public Policy, 39, December, 95–214.
Walsh, C. E. (1997) Monetary Theory and Policy, MIT Press, Chicago.
Walsh, C.E. (2001) Teaching inflation targeting: an analysis for
In this paper an optimal policy rule for the central bank to intermediate macro. Forthcoming in Journal of Economic
achieve its monetary policy goals has been presented, Education. (LATEST INFO??).