Monetary Policy Rules in Colombia
| Autor | Raquel Bernal |
| Páginas | 37-53 |
#
!""#
DESARROLLO Y SOCIEDAD
!
Monetary Policy Rules in Colombia*
D?CC
Resumen
Este estudio reporta estimaciones de una función de reacción de política
monetaria para Colombia durante el período 1991 a 1999. Los resultados
indican que durante este período el Banco Central adoptó una regla implí-
cita cuyo objetivo era la tasa de inflación. La evidencia también sugiere
que el Banco Central respondió a la inflación anticipada en contraposición
a la inflación rezagada. De acuerdo con los resultados, aún bajo la existen-
cia de una banda cambiaria cuyo objetivo era servir de ancla nominal para
la política monetaria, el Banco Central tuvo espacio para movimientos in-
dependientes de la tasa de interés siempre y cuando la tasa de cambio no
se encontrara en alguno de los dos extremos de la banda.
Clasificación JEL: E0, E4, E5.
Palabras clave: Función de reacción de política monetaria, tasa de infla-
ción, tasa de interés, regla de Taylor.
* I want to acknowledge helpful comments from Kenneth Kuttner, Mark Gertler, Sergio
Clavijo, Arturo Galindo and an anonymous referee.
** New York University. Department of Economics (e-mail: rb459@nyu.edu). First version
of this paper: February 2000.
ISSN 1900-7760
(Edición Electrónica)
"
)##"
"*+
Introduction
This paper characterizes empirically how the Central Bank of Colombia
has conducted monetary policy since 1991 when the new political constitution
declared complete autonomy of the Bank in policy management with the
sole objective of controlling the growth of prices in the economy. The
announcement of inflation objectives began in Colombia in the early 1990s.
In 1991, the new constitution established the independence of the Banco
de la República and mandated that the design and execution of monetary,
exchange rate and credit policies was the exclusive responsibility of its
Board of Directors. It also stated that the Central Bank needed to “preser-
ve the purchasing power of the currency” (article 373), and in 1992 a new
law (Law No. 32) mandated that the Board of Directors announced a
quantitative inflation objective each year1.
There are two main issues that motivate this research. First, after more than
a decade of high and persistent inflation (an average of 24.9% between
1975 and 1992), Colombia has experienced a considerable decline in price
growth during the past eight years when inflation has finally decreased below
20% and has even reached single digit numbers in 1999. This period coin-
cides with that for which the Central Bank has been independent with the
sole concern of reducing price growth. It is worthwhile then to assess the
performance of the Central Bank during this period and bring forward some
of the most important lessons of this experience for future policy-making in
Colombia.
Second, given that for most of the period under analysis the country used
an exchange rate band, it can be interesting to think of the monetary policy
during the last years as one in which the interest rate is an operating instrument
while the exchange rate is an intermediate target. The estimation of a policy
reaction function can be useful to evaluate whether the Central Bank still
had space for independent movements in the target interest rate, so long as
the exchange rate was not exactly at either edge of the band.
In this paper, I estimate a monetary policy reaction function for Colombia
during the period between September 1991 and October 1999.
1See Uribe D. et al., 1999.
#
!""#
DESARROLLO Y SOCIEDAD
#
Section 1 below presents an estimable policy reaction function. The baseline
specification has a central bank adjusting the nominal short term interest
rate in response to the gaps between expected inflation and output and
their respective targets. The model corresponds to the kind presented by
Clarida, Gali and Gertler (1997) and employed in their study of monetary
policy in several countries including the U.S., Germany and Japan. It is
essentially a forward looking version of the simple backward looking reaction
function popularized by Taylor (1993).
Additionally, I present several alternatives to the baseline specification in
which the central bank responds to variables other than inflation and output.
This can be particularly relevant given that the country faced external
constraints on monetary policy imposed primarily by the target zone. One
of these alternatives allows to test the forward looking versus the backward
looking specifications of the reaction functions.
Section 2 presents the results of the analysis for Colombia. Overall, the
baseline specification of the reaction function does a good job in
characterizing monetary policy for the country after 1991. The kind of
rule that emerges is what one might call “soft-hearted” inflation targeting:
In response to a rise in expected inflation relative to target, the Central
Bank raises nominal rates sufficiently to push up real rates. The estimated
rule thus implies a clear focus on controlling inflation. At the same time,
there is a modest pure stabilization component: holding constant
expected inflation, the Central Bank adjusts rates in response to the
state of output.
Concluding remarks are in section 3.
I. Specification of a Monetary Policy Reaction Function
The basic framework pertains to a central bank that has at least some
degree of autonomy over its monetary policy. The starting point is the
observation that, for most central banks, the main operating instrument
of monetary policy is a short term interest rate. Typically, the instrument
is an interbank lending rate for overnight loans. The empirical policy
reaction function developed by Clarida, Gali and Gertler (1997)
characterizes how central banks choose the level of the short term rate
from period to period.
$
)##"
"*+
To make sense of this class of policy reaction functions, the analysis appeals
to existence of temporary nominal wage and price rigidities. With nominal
rigidities, monetary policy affects real activity in the short run. By varying
the nominal rate, a central bank can effectively vary the real interest rate
and the real exchange rate. The process of imperfect wage and price
adjustment gives rise to a positive short run relationship between output
and inflation. This implies, for example, that reducing inflation may require
a period of output reduction, depending on the degree of nominal
stickiness. In this kind of setup, is possible to envision a central bank
choosing the course of short term interest rates.
Assume that each period the central bank has a target for the nominal short
term interest *
t
r. In the baseline case, the target will depend on both expected
inflation and output:
()
[
]
()
[
]
** ttttnt
*
tyyEErr −Ω+−Ω+= +
γ
ππ
β
(2.1)
where r is the long run equilibrium nominal rate, nt+
π
is the rate of inflation
between periods t and t + n, yt is real output, and π* and *
t
yare respec-
tive bliss points for inflation and output. *
t
y is given by potential output,
defined as the level that would arise if wages and prices were perfectly
flexible. E is the expectation operator and Ωt is the information available
to the central bank at the time it sets interest rates. Equation (2.1) implies
that the short term interest rate adjusts in response to deviations of inflation
from target, as well as in response to deviations of output from the flexi-
ble price level of output.
The nominal interest rate equilibrium level is defined as:
*
π
+= rrr (2.2)
and
()
tnttt Errr Ω−= +
π
where rrt is the real interest rate. If we combine these equations we obtain:
()
()
[
]
()
[
]
*** 1 ttttntt yyEErrrr −Ω+−Ω−+= +
γ
ππ
β
(2.3)
#
!""#
DESARROLLO Y SOCIEDAD
where rr is the long run equilibrium real rate of interest. According
to equation (2.3), the target real rate adjusts relative to its natural
rate in response to departures of either expected inflation or output
from their respective targets. The magnitude of the parameter
β
is
key to this analysis. If
β
> 1 the target real rate adjusts to stabilize
inflation. With
β
< 1, the monetary policy will instead accommodate
changes in inflation: even if the central bank raises the nominal rate in
response to an expected rise in inflation, for example, it does not
increase it sufficiently to keep the real rate from declining. The
estimated magnitude of the parameter
β
provides an important
yardstick for evaluating a central bank’s policy rule.
This target policy is a generalization of the type of the simple interest
rate rule proposed by Taylor (1993) and Henderson and McKibbon
(1993) among others. This rule, has the central bank respond to lagged
inflation as opposed to expected future inflation. The specification in
Clarida, Gali and Gertler (1997) however, nests the “Taylor” rule: if
either lagged inflation or a linear combination of lagged inflation and the
output gap provides a sufficient statistic for inflation, then the
specification reduces to the simple Taylor rule. The general specification
has several virtues: (i) by explicitly incorporating expected inflation
makes it easier to disentangle the link between the estimated coefficients
and central bank objectives and (ii) by having the central bank respond
to forecasts of inflation and output, this general specification
incorporates a realistic feature of policy-making according to which
central banks consider a broad array of information.
A simple equation like (2.1), however, does not capture the tendency of
central banks to smooth changes in interest rates2. In order to capture
these factors, it is simply assumed that the actual rate partially adjusts to the
target as follows:
()
tt
*
tt rrr
νρρ
++−= −1
1 (2.4)
where the parameter
ρ
∈¸ [0, 1] captures the degree of interest rate
smoothing, and vt is assumed to be serially uncorrelated. vt could reflect a
2See e.g. Goodfriend (1991).
)##"
"*+
pure random component to policy, or could arise because the central bank
imperfectly forecasts idiosyncratic reserve demand, and for some reason,
does not instantly supply reserves to offset the shock.
We can rewrite equation (2.1) in the following way:
()()
tttntt xEEr Ω+Ω+= +
*
γ
π
β
α
(2.5)
*
β
πα
+≡ r
*
ttt yyx −=
Further, we can combine equations (2.4) and (2.5) to obtain:
()
()(){}
tttttntt vrxEEr ++Ω+Ω+−= −+ 1
1
ρ
γ
π
β
α
ρ
(2.6)
Finally, we eliminate the unobserved forecast variables from this expression
by rewriting the policy rule in terms of realized variables as follows:
()() ()
tttntt rxr
εργρβπραρ
++−++−= −+ 1
1–1 1 (2.7)
where the error term
ε
t depends on vt and the forecast errors of inflation
and output gap:
()
()
[]
()
[]
{}
tttttntntt vxExE +Ω−+Ω−−−= ++ 1
γ
ππ
β
ρ
ε
(2.8)
Given that we assume rational expectations, we know that the forecast
errors will be uncorrelated with any information at time t. Hence we can
define ut as a collection of variables observable to the central bank at t, but
orthogonal to to vt3. Possible elements of ut include any lagged variables
that help forecast inflation and output, as well as any contemporaneous
variables that are uncorrelated with the current interest rate shock vt. Then,
3The econometric approach relies on the assumption that, within the short sample, short
term interest rates and inflation are I(0). Standard Dickey-Fuller and Phillips-Perron tests
of the null that the short term interest rate is I(1) are rejected in favor of the alternative of
stationarity. In the case of the inflation rate there is less evidence against the null that this
series is I(1). However, we know that the Dickey-Fuller test has low power against the
alternative of stationarity for short samples, which is our case.
#
!""#
DESARROLLO Y SOCIEDAD
since E (εt|ut) = 0, equation (2.7) implies the following set of orthogonality
conditions that will be used for estimation:
()() ()
[]
0 1–1 1 1=−−−−−− −+ tttntt urxrE
ρ
γ
ρ
β
π
ρ
α
ρ
(2.9)
To estimate the parameter vector [
β, γ, ρ, α
] I use generalized method
of moments4. In the baseline case the instrument set ut includes lagged
values of output, inflation, interest rates, and the change of the log of
the real exchange rate. Each of these variables is potentially useful for
forecasting inflation and output and is exogenous with respect to the
interest rate shock, given our identifying assumptions. Since the potential
instrument set, and hence the number of orthogonality conditions,
exceeds the parameter vector, the model is overidentified, in which case
it is straightforward to test the over-identifying restrictrions (Hansen
(1982)). Under the null hypothesis of this test, the central bank adjusts
the interest rate each period so that (2.6) holds, with the expectations
on the right hand side based on all the relevant information available to
policymakers at that time. Under the assumptions of the model, this
implies the existence of values for [
β, γ, ρ, α
] such that the implied
residual
ε
t is orthogonal to the variables in the information set Ωt. Under
the alternative, however, the central bank adjusts the interest rate in
response to changes in some current and/or lagged variables, but not
necessarily in connection with the information that those changes contain
about future inflation and output. In this case, some relevant “explanatory
variables” might have been omitted from the interest rate equation (2.7).
To the extent that some of those variables are correlated with ut, the set
of orthogonality conditions will be violated, which would lead to a
statistical rejection of the model.
Using the parameter estimates of
α
and
β
, it is possible to recover an estimate
of the central bank’s target inflation rate π*. While we cannot separately
4The composite disturbance term of the model has an MA(n-1) structure (Hansen and
Hodrick (1980)). In this case the GMM estimator of the parameter vector is a two-step
nonlinear two-stage least squares estimator (Hansen (1982)) when the model is
overidentified. In the first step, we use traditional non-linear two-stage least squares to
obtain an initial estimate of the parameters. Then these initial parameter estimates are
used to construct an optimal weighting matrix.
)##"
"*+
identify π* and rr , the model provides a relation between the two variables
that is conditional on
α
and
β
. Given that *
β
πα
−≡ r and
()
** 1 ,
πβαπ
−+≡+= rrrrr which implies:
1
*
−
−
=
β
α
π
rr (2.10)
The sample average real rate can be used to provide an estimate of rr .
Then it is possible to construct an estimate of π*. Recall that π* is defined
as the long run inflation target. In the case of Colombia, however, the
Central Bank set decreasing annual inflation targets for most part of the
period and starting in 2001 the Bank set a multi-year target and announced
a long term inflation target. In this case, the existence of a unique long
term inflation could be justified by the assumption that there is an underlying
long term inflation target in spite of the fact that the Bank sets several
short term inflation targets or if one thinks of the long term inflation target
as an average of short term intermediate inflation targets. It would be
feasible to estimate the model assuming that the Central Bank sets more
than one inflation target, π*, over the period. However, I do not expect
the results to be very sensitive to such a change since on average one
would expect the underlying long term inflation target to roughly coincide
with the average of the intermediate short term goals set by the Central
Bank over that period.
For the Central Bank of Colombia I estimate the baseline specification
for the period for which there was at least some degree of autonomy
over domestic monetary policy. It is possible that there may be other
important factors that influence interest rate setting besides those
captured in the baseline model as I discussed in the introduction. In this
case, while the Central Bank did not completely sacrifice monetary
control, it may have pursued policies to maintain exchange rates within
reasonable bounds. Exchange rates may thus have influenced policy,
independently of the information they contain about inflation and output.
In addition, it is possible that the Central Bank could have paid attention
to monetary aggregates. To account for these possibilities I consider a
number of simple alternatives to the baseline policy. Let zt denote a
variable besides inflation and output that may potentially influence rate
setting (independently of its use for forecasting). For each alternative
#
!""#
DESARROLLO Y SOCIEDAD
specification, we then replace the relation for the target given by equation
(2.5) with the following
()()
[]
tttttntt zExEEr Ω+Ω+Ω+= +
ξ
γ
π
β
α
*(2.11)
The alternative specification is estimated in the same fashion as the
baseline, except that the parameter vector is expanded to include the
coeffcient ξ on the additional variable zt, and expand our instrument list
to include lagged values of that variable. After doing so, it is
straightforward to evaluate the quantitative importance of zt on policy.
The variables I consider include: real exchange rate, money supply and
lagged inflation. By including the latter I obtain a direct test of the forward
looking specification of the policy rule versus the “backward looking”
specification of the Taylor rule.
In estimating (2.9) or alternatively the set of orthogonality conditions
implied by (2.11) I consider the horizon of the inflation forecast that
enters the reaction function as being one year, i.e. n = 12. According to
Clarida, Gali and Gertler (1997), it is reasonable to believe that policy-
makers are unconcerned about the month to month variation in inflation
and instead are more concerned about medium and longer term trends.
Hence, a year ahead forecast seems to be a good indicator of a medium
term trend in inflation.
II. Monetary Policy Rules for Colombia after 1991
I now proceed to estimate a monetary policy reaction function for the Cen-
tral Bank of Colombia. The bank has been virtually autonomous over its
domestic monetary policy since 1991 when the new political constitution
declared complete autonomy of the Central Bank and gave them the unique
objective of controlling inflation. Given that the country used an exchange
rate band during most of this period, the baseline specification according to
which policy responds purely to domestic macroeconomic conditions has
to be interpreted with caution. For several episodes, particularly during
1998 and 1999 the defense of the exchange band was associated with
persistent changes in the inter-bank lending rate independently of changes
in inflation and the output gap. For these reasons, an alternative specification
that includes the real exchange rate as an additional determinant of the
)##"
"*+
short term interest rate might be more relevant to a country that faced this
type of external constraints5.
Figure 1 plots the rate of consumer price inflation6 and the short term interest
rate (the inter-bank lending rate) for the period between 1989 and 1999.
Note first the high and persistent rate of inflation for the first part of the
period, with rates even above 30% at the end of 1990 and beginning of
1991. In 1992 we can observe a clear pattern of disinflation up to 1999
when the inflation rate was around 9.2%.
5Clavijo (2002) mentions that the inclusion of the real exchange rate in a Taylor rule is not
the best approach to follow given the problems associated to the uncertainty of the long-
term purchasing power parity. Instead, he uses the (unconvered) interest rate parity
condition and argues that its effect on net international reserves translates into changes in
the relation between monetary aggregates and domestic interest rates.
6Frequently, the Central Bank responded to deviations in the underlying (“basic”) inflation rate
with respect to the target. In other words, monetary policy did not react (or only subtly) to
supply shocks unless they were significantly reflected in higher expectations of inflation. For
these reason, one could estimate an alternative specification of the model with different
measures of “basic” inflation rather than the percentage change in the consumer price index.
#
!""#
DESARROLLO Y SOCIEDAD
!
On the other hand, we can observe a change in the pattern of behavior of
short term interest rate that occurs roughly around 1993. Prior to this date,
the central bank kept short term interest rate at or below the rate of inflation.
Real short term rates accordingly hovered around zero and below. During
1992 and 1993 however, real as well as nominal short term rates moved
up significantly. What can be key is not exactly the secular behavior of
short term rates, but rather a sort of co-movement with inflation that we
can observe during this period. One possible explanation that can arise is
that the Central Bank shifted the monetary policy to raise short term real
rates to offset inflationary pressures. This conjecture is investigated formally
by estimating the policy reaction function described in section 1.
The starting date for the estimations is September 1991. In July 7th 1991
the new political constitution was signed. Since that day, the Central Bank
has been autonomous to conduct monetary policy in the country.
Nevertheless I allow for a two-month adjustment in which the institutions
restructured according to their new functions and objectives. For this reason
the sample starts in September 1991 and ends in August 1999 which is the
last available data. I use the consumer price index to measure inflation and
an index of industrial production to measure output7. To obtain a measure
of the output gap, I detrend the log of industrial production using a quadratic
trend8. The interest rate is the interbank lending rate9.
I first estimate the baseline specification for policy, given by equation (2.6).
The instrument set includes 1-6, 9, 12 lagged values of: the output gap yt,
inflation πt, the interbank lending rate rt and the log difference of the real
exchange rate, qt.
The top line of Table 1 reports the results for the baseline specification. The
key result is the estimate of the coefficient on the inflation gap,
β
: 1.34 with
a standard error of 0.18. A rise in expected annual inflation of one percent
7The industrial production series was deseasonalized to avoid having a seasonal component
in the measure of output gap.
8An alternative measure of the output gap that uses a Hodrick-Prescott filter yields very
similar estimates. Results are available upon request.
9It is worth mentioning, however, that starting in 1995 the repo-rate was instituted as one
of the main instruments of monetary policy.
"
)##"
"*+
induces the Central Bank to raise real rates by 34 basis points. Because
β
is significantly greater than one, the prediction that the Central Bank raises
real rates in response to inflationary pressures is statistically significant.
Another interesting result is that the estimate of the coefficient on the output
gap is positive and also statistically significant: 0.19 with a standard error of
0.06. Thus, holding constant expected inflation, a one percent rise in the
output gap induces the Central Bank to increase nominal (and thus real)
rates by 19 basis points. The GMM estimation procedure yields
(asymptotically) correct standard errors, and thus allows to confirm the
statistical significance results.
'!( )* + , (
ββ
ββ
βγγ
γγ
γρρ
ρρ
ραα
αα
αξξ
ξξ
ξ
Baseline 1.34 0.19 0.10 0.59 ---
(0.18) (0.06) (0.06) (0.07)
Adding:
Lagged Inflation11.14 0.08 0.005 -0.41 -0.32
(0.23) (0.04) (0.02) (0.04) (0.27)
Money Supply21.24 0.10 0.06 -0.53 -0.35
(0.15) (0.04) (0.05) (0.06) (0.18)
Real Exchange Rate31.44 0.24 0.11 -0.34 0.002
(0.17) (0.03) (0.05) (0.04) (0.00)
Test of Overidentifying Restrictions for:
Baseline Specification J = 23.1 with p-value = 0.844
Specification 1 J = 24.2 with p-value = 0.933
Specification 2 J = 27.3 with p-value = 0.949
Specification 3 J = 27.2 with p-value = 0.921
The sample is 1991:9 - 1999:8.The instruments are yt-1 ...yt-6 ,yt-9, yt-12, p t-1 ...p t-6 ,p t-9 , p t-12, rt-
1 ...r t-6 , r t-9 , r t-12 ,z t-1 ...z t-6 , z t-9 , z t-12 , q t-1 ...q t-6 , q t-9 , q t-12 where q is the change in the log
of the real exchange rate. Estimates are obtained by GMM with correction for MA(12)
autocorrelation. Optimal weighting matrix obtained from first step two-stage least squares parameter
estimates.
1. p t-12
2. M1 money growth over last 3 months.
3. Lagged levels of real exchange rate also included in instrument list.
Finally, the J-statistic implies that we do not reject the overidentifying
restrictions of the baseline model. These results are in clear contrast with
those reported by Clavijo (2002). According to the author, estimates of
#
!""#
DESARROLLO Y SOCIEDAD
#
the Taylor rule are not satisfactory with both, monetary and output gaps,
turning out insignificant and with unexpected signs. Only the interest rate
parity condition appears to be significant. These results, however, must be
interpreted carefully given that as can be observed from equation (2.7), the
error term
ε
t depends on vt (interest rate shock) and the forecast errors of
inflation and output gap. This means that ordinary least squares estimates
of the Taylor equation are inconsistent.
The sample average real rate -which I take as the estimate of the long run
real rate- is 4.55. Using this estimate, I obtain an estimate of the long run
inflation target, π* of 11.4 which roughly corresponds to the average of
annual objectives set by the Central Bank since the date in which it is
considered independent which equals approximately 13%.
Next, I consider alternatives to the baseline specification. First, I allow
lagged inflation to enter the reaction function, along with expected inflation
and output. Lagged inflation is not statistically significant and the point estimate
for the associated coefficient has the wrong sign. Furthermore, this alternative
specification does not perceptibly change the estimate of the coefficient
β
although the estimate of γ is reduced to half. Taken together, the results
suggest that the backward-looking specification can be rejected in favor of
the forward –looking one.
The money supply is also unimportant. In this case, I include an average of
the past three months money (M1) growth in the reaction function. The
variable is not significant at the 5 percent level. The other key coefficients in
the reaction function are virtually unchanged.
Finally I consider the level of the real exchange rate as an external constraint
to monetary policy. Each variable does enter significantly and with the right
sign, but the quantitative effect of the real exchange rate is considerably
small. Once again, the estimates of the slope coefficients are virtually the
same as in the baseline case. The results indicate that a 10 percent real
depreciation induces a 2 basis point increase in the short term interest rate.
To gain some feel of how well does the baseline specification in explaining
the behavior of the Central Bank, I plot the implied target rate versus the
actual rate. I include 1990 in the sample just for comparison purposes.
This means, that I compare the implied target using the post-1991 rule to
$
)##"
"*+
the pre-1991 data as well as to the post-1991 data. This might give an idea
of the policy shift after independence was granted to the bank.
10 Note that I use the target rate as opposed to the fitted rate (which would include the lagged
interest rate).
Figure 2 shows these results10. We can observe, that starting in 1992 the
target rate implied by the estimated model nicely tracks the behavior of
actual rates in Colombia. We can observe, still, temporary deviations. Note
for example that for most of 1997 the target rate is above the actual. During
this year the exchange rate remained for most of the time at the bottom of
the band, hence it is possible that the Central Bank did not need to be as
tight. The comparison with pre-1991 is clear. The post-1991 rule implies a
target rate much higher than the actual rate.
#
!""#
DESARROLLO Y SOCIEDAD
III. Concluding Remarks
The results presented above show that since 1991 the Colombian Central
Bank has pursued what can be called a “soft-hearted” inflation targeting: In
response to a rise in expected inflation relative to target, the Central Bank
raises nominal rates sufficiently to push up real rates. This behavior is
statistically significant and quantitatively important. However, there is a
modest pure stabilization component to the rule, holding constant expected
inflation, the Central Bank adjusts rates in response to the position of output
relative to trend, but not by a significant amount. The primary focus of
monetary policy appears to be on inflation. When including the real exchange
rate in the specification, the coefficient turns out to be significant but
quantitatively small. The inclusion of this variable, does not change the other
slopes considerably.
This kind of simple rule appears to set the economy on a course for stable
long term inflation using relatively little knowledge about the economy.
Because the rule is simple for the private sector, it is conducive to building
and maintaining credibility.
On the other hand, even with the existence of the exchange rate band,
it can be said that the Central Bank still had scope for independent
movements in the target interest rate, so long as the exchange rate was
not right at either edge of the band. In the cases in which the exchange
rate was at the bottom edge of the band, the bank could loosen its
policy temporarily.
In this paper, I have assumed that the Central Bank sets a unique long run
inflation target, π*, over the period. However, the Colombian Central
Bank set annual inflation goals over most part of the period and only in
2001 did it set multi-year targets and announced a long term inflation
target. Hence it would be interesting to extend the model in the future to
allow for the possibility of several inflation targets and assess whether the
results are sensitive to a change in this particular assumption. However, I
do not expect the results to vary significantly since on average one would
expect the underlying long term inflation target to roughly coincide with
the average of the intermediate short term goals set by the Central Bank
over the period.
)##"
"*+
References
Ball, L. (1997 ). “Efficient Rules for Monetary Stability”, NBERW.P. no.
5952, March.
Clarida, Richard, Gali, Jordi and Gertler, Mark, forthcoming, “Monetary
Policy Rules and Macroeconomic Stability: Evidence and Some
Theory”, Quarterly Journal of Economics.
____. (1998). “Monetary Policy Rules in Practice: Some International
Evidence”, European Economic Review, June.
Clavijo, S., (2002). “Monetary and Exchange Rate Policies in Colombia:
Progress and Challenges (1991-2002)”, Manuscript, Banco de la
República, Bogotá, Colombia.
Goodfriend, M. (1991). “Interest Rate Smoothing and the Conduct of
Monetary Policy”, Carnegie-Rochester Conference on Public Policy,
Spring, pp. 7-30.
Hansen, L. (1982). “Large Sample Properties of Generalized Method of
Moments Estimators”, Econometrica, 50, , 1029-1054.
Hansen, L. and Hodrick, R. (1980). “Forward Exhange Rates as Optimal
Predictors of Future Spot Rates”, Journal of Political Economy
88, October, pp. 829-853.
Henderson, D. and Mckibbon, W. (1993). “An Assessment of Some Basic
Monetary Policy Regime Pairs”, Chapter 2 in Evaluating Policy
Regimes: New Research in Macroeconomics, R. Bryant, ed. Was-
hington: Brookings.
King, M. (1996). “How Should Central Banks Reduce Inflation? Con-
ceptual Issues”, in Achieving Price Stability, Kansas City: Federal
Bank of Cansas City.
Obstfeld, M. and Rogo, K. (1995). “The Mirage of Fixed Exchange Rates”,
Journal of Economic Perspectives, (4) Fall, pp. 75-96.
#
!""#
DESARROLLO Y SOCIEDAD
Svensson, L. (1996). “Inflation Forecast Targeting: Implementing and
Monitoring Inflation Tragets”, NBER W.P. no. 5797, October.
Taylor, J. (1993). “Discretion Versus Policy Rules in Practice”, Carnegie-
Rochester Conference on Public Policy, 39, pp. 195-214.
Uribe, J., Gómez, J., and Vargas, H. (1999). “Strategic and Operational
Issues in Adopting Inflation Targeting in Colombia”, Manuscript,
Banco de la República, Bogotá, Colombia.
