The Impact of Exchange Rate Regime on Interest Rates in ... · Exchange Rate Regimes and Interest Rates in Latin America 93 (1979) model to take into account emerging countries’

CUADERNOS DE ECONOMÍA, VOL. 47 (MAYO), PP. 91-124, 2010

* I thank Eric Girardin as well as an anonymous referee for comments and suggestions which were very helpful in improving the present paper. Email: [email protected]

The Impact of Exchange Rate Regime on Interest Rates in Latin America*

Caroline DuburcqUniversité de la Méditerranée, Aix-Marseille II

We develop a theoretical framework to study the impact of the exchange rate regime in the interest rate determination. Using VECM, we assess the role of both domestic conditions and US factors in the determination of eight Latin-American countries’ interest rates between February 1998 and April 2009. Three countries have hard-peg while the remaining five follow alternative regimes. The long and short-run determinants of domestic rates as well as an impulse response analysis prove that economies with rigidly-fixed exchange rates do not bear a loss of monetary autonomy substantially higher than that of floating-rate economies, with the exception of Brazil.

JEL: E�3, F31, C32Keywords: Interest Rate Determination, Exchange Rate Regime, Vector Error Correction Models

1. Introduction

Choosing an exchange rate regime is a fundamental macroeconomic policy decision, especially for small open economies. The choice to adopt fixed exchange rates or not, may determine policy options and/or the ability to maintain open capital markets. In this paper, we test a basic proposition of international macroeconomics, the notion of the open-economy trilemma (Mundell, 1963), which implies that countries cannot have fixed exchange rates, domestic monetary autonomy, and open capital markets all at once, but can only pursue two of these options.

We can explain the behavior of short-term interest rates with two different approaches (Barassi et al, 2005). Interest rates can either be treated as analogous to other asset prices, in which case their movements are interpreted as being

92 Cuadernos de Economía Vol. �7 (Mayo) 2010

determined by financial flows in profit-seeking capital markets, giving rise to a set of arbitrage conditions such as uncovered interest parity. Or they can be viewed as policy instruments and are then determined by decisions aiming at a policy objective such as an exchange rate or an inflation target. There is a long standing literature on the latter (Clarida et al., 1998; 1999; 2001; Adam et al., 2005). In this paper, our objective is different as we study the interest rate behavior in the long run, taking into account both internal and external determinants and making a systematic link with exchange rate policy.

Although monetary independence has been at the heart of the debate on exchange rate regimes, empirical evidence on this issue is still mixed. Shambaugh (200�) uses a sample of 100 industrialized and developing countries from 1973 to 2000 and finds that fixed countries’ interest rates strongly follow the base country interest rate changes. Obstfeld et al. (200�; 2005) also find that the interest rates of floating-rate economies show far less connection to the base country’s interest rates than hard-peg countries. Borensztein et al. (2001) also report evidence consistent with the traditional view of more monetary independence for flexible-rate countries. On the opposite, some papers find evidence consistent with the alternative view, namely, the more firmly pegged is a country to the dollar, the smaller its reaction to changes in US interest rates. This is the case of Frankel (1999) who studies the effects of nominal interest rate fluctuations in the United States on domestic rates in Argentina, Brazil, Hong Kong, Mexico and Panama between 1993 and 1998 and finds that this effect is significantly larger in economies with no credible commitment to a fixed exchange rate parity (Brazil and Mexico) than in countries with « truly fixed » regimes (Argentina, Hong Kong and Panama). Hausmann et al. (1999) use a monthly panel of 11 Latin American countries between 1960 and 1998 and conduct a similar exercise. They conclude that the effect of US interest rate on domestic rates is 25% lower in fixed-exchange-rates countries.

The “fear of floating” literature, initiated by Calvo and Reinhardt (2002), states that only large countries can benefit, or choose to benefit, from an independent monetary policy, as many declared floating-rate countries de facto limit exchange rate flexibility and may not have or use the autonomy attributed to floating rates. For Frankel et al. (2002), fixing the exchange rate does not generate a loss of monetary flexibility, as most countries would not have freedom even if they had floating rates. This is in line with the growing body of literature that states that emerging markets economies, prone to important changes in international investors’ confidence, cannot benefit from the use of the interest rate instrument and that would actually be worse to let them that possibility. The question we pose here is whether the exchange rate regime influences the extent to which domestic short term interest rates are caused by internal and/or external factors.

This paper is an extension of the existing literature in that we not only look at relationships between domestic and base country’s interest rates but we allow for a set of both internal and external factors as possible determinants of local interest rates in the long and short run while making a systematic link with the exchange rate regime. By external factors, we refer to US variables, as we are looking at Latin American countries. We develop a revised version of Frankel’s

Exchange Rate Regimes and Interest Rates in Latin America 93

(1979) model to take into account emerging countries’ specificities. Namely, we allow for imperfect substitutability of domestic and foreign assets. The long-run history of sovereign debt in Latin America indicates that bonds are frequently traded at market values showing substantial levels of risk premium. We also model currency substitution in the domestic money demand specification as it is a very significant phenomenon in Latin America.

Using vector error correction models, we assess empirically the role of both domestic conditions and US factors in the determination of eight Latin American countries’ interest rates, with monthly data over February 1998 through April 2009. Three are hard-peg countries, the remaining five have flexible or intermediate exchange rate regimes as calculated with our update of the Levy-Yeyati and Sturzenegger (2005) de facto classification method, based on data on exchange rates and reserves. The long and short-run determinants of domestic interest rates as well as an analysis of impulse response functions prove that economies with rigidly-fixed exchange rates do not bear a loss of monetary autonomy substantially higher than that of floating-rate economies, with the exception of Brazil, the region’s largest country and the only floating-rate-economy of our sample that proves to benefit from monetary freedom.

The rest of paper is organized as follows. The next section provides a brief description of the macroeconomic framework, Section 3 describes the data and econometric model, Section � presents the results and finally in Section 5 we conclude.

2. Conceptual Framework

Based on Frankel’s (1979) model, we develop a simple macroeconomic framework to study interest rate determination. Our first assumption is an interest rates parity condition distorted by a risk premium as we are considering emerging market economies:

(1) it = it*+ xt + rt

where it is the domestic nominal interest rate; it* is the foreign nominal interest

rate; xt is the expected rate of depreciation of the domestic currency quoted as the number of units of domestic currency per unit of foreign currency; and rt is a time-varying risk premium.

Note that in equation (1) we do not assume efficient markets in which sovereign bonds would be perfect substitutes. The long-run history of sovereign debt in Latin America indicates that bonds are frequently traded at market values showing substantial levels of risk premium. The empirical literature (Alper et al., 2007) on the uncovered interest parity condition reveals that emerging countries deserve a special treatment due to specific macroeconomic conditions including incomplete institutional reforms, weaker macroeconomic fundamentals, and shallow financial markets.

9� Cuadernos de Economía Vol. �7 (Mayo) 2010

As in Dornbusch’s (1976) model, we distinguish between the long-run exchange rate, to which the economy will ultimately converge, and the current exchange rate. Denoting the logarithms of the current and long-run exchange rates by et and e , respectively, we assume that:

(2) x e et t= −( )θ

Equation (2) states that the expected rate of depreciation of the spot rate is proportional to the discrepancy between the long-run rate and the current spot rate. The long-run exchange rate is assumed known, and an expression for it will be developed below. We assume purchasing power parity holds in the long run:

(3) e p p= − *

where p and p* are defined as the logarithms of the equilibrium price levels at home and abroad, respectively.

We assume a domestic money demand specification that takes into account the most significant phenomenon in Latin America, namely currency substitution. Based on the long standing literature on currency substitution (Miles, 1978; Arize, 199�; de Freitas and Veiga, 2006), we consider that the conventional money demand equation must be augmented with the exchange rate:

(�) m p y i et t t t t= + − −ϕ λ ψ

where mt, pt and yt are defined as the logarithms of the nominal quantity of money, the price level and the real income respectively. A conventional money demand function holds abroad:

(5) m p y it t t t* * * *= + −δ λ

As in Frankel’s model, we assume that the interest rate semi-elasticities of money demands are the same for the domestic and foreign countries. Let us take the difference between the two equations (�) and (5):

(6) m m p p y y i i et t t t t t t t t− = − + − − − −* * * *( )ϕ δ λ ψ

Using bars to denote equilibrium values, and remembering that in the long run, when e e i i= − =, * r, we obtain:

(7) e m m y y=−

− − + +

1

1 ψϕ δ λr* *

Substituting (7) into (1), and assuming, as in Frankel’s (1979) model, that the current equilibrium money supplies, income levels and risk premium are


given by their current actual levels, we obtain a complete equation of interest rate determination:

(8) i i e m m y yt t t t t t t= + −−

−

+

−−

−+

− −* * *θθ

ψθφ

ψθδ

ψψ θλ

1 1 1

1

1−−ψrt

Simplifying with αθ

ψβ

θφψ

ζθδ

ψγ

ψ θλψ

=−

=−

=−

=− −

−1 1 1

1

1, , y

we obtain:

(9) i i e m m y yt t t t t t t t= + − −

+ − +* * *θ α β ζ γr

The domestic interest rate is positively related to the foreign interest rate, the exchange rate, the external money supply, the domestic level of income and the risk premium and negatively related to the domestic money supply and the foreign level of income. This relation is tested empirically for a set of eight Latin American countries.

3. Data and Econometric Methodology

The monthly data set runs from February 1998 to April 2009. We study the eight Latin American countries for which the JP Morgan Emerging Market Bond Index plus (EMBI+) spread is reported, namely Argentina, Brazil, Colombia, Ecuador, Mexico, Panama, Peru and Venezuela. The EMBI+ is a US dollar emerging markets debt benchmark while the EMBI+ spread, commonly known as sovereign spread, measures the credit risk premium over US Treasury bonds.

Within these countries, we have three hard-peg experiences: Panama on the whole sample, Ecuador as of January 2000 and Argentina between January 1998 and December 2001. The remaining five countries follow either flexible or intermediate exchange rate regimes as calculated with our update of the Levy-Yeyati and Sturzenegger (2005) de facto classification method. They are used as control countries. We also look at the case of Argentina restricted to the floating period as a sixth experience of control country. Our sample excludes hyperinflation periods which increase the probability that the domestic and the US time series have the same integration properties. The EMBI data has been obtained from JP Morgan and stands for the risk premium. As a proxy to measure monetary policy, we use a short-term interest-rate, the 90-day interbank market rate when available or the deposit 90-180 day rate as an alternative. Data has mainly been extracted from the IMF’s International Financial Statistics (IFS). We use the nominal exchange rate (expressed as national currency per US dollars), a M1 index, the consumer price index and an industrial production index. More details on the data used and samples are given in the Appendix, Table A1.


We estimate our model of interest rate determination for each country. We first check the order of integration of the data using ADF1, Phillips-Perron (1988), KPSS2 and Ng-Perron (2001) unit root tests. All series are integrated of order one, results of these tests are presented in the Appendix, Table A2. We then conduct the Johansen (1988) and Johansen and Juselius (1990) cointegration procedure to test for the presence of cointegrating vectors between the domestic interest rate, a set of internal variables and a set of external variables. The procedure is based on the maximum likelihood estimation of the vector error correction model (VECM):

(10) ∆zt = Π zt−1 + Γ1 ∆zt−1 + …. + Γp−1 ∆zt−p+1 + κ + ut

where the matrix Γ captures the short-run aspects of the relationships between the elements of zt and the matrix Π reflects the long-run information. The rank of Π, denoted by r, determines the number of cointegrating relations. The matrix Π can be decomposed into two matrices, α and β where Π = αβ’. The weights, also called the error coefficients, are contained in matrix α that forces the series back towards their underlying equilibrium relations while the cointegrating vectors are contained in matrix β that gives the underlying long-term relations.

According to our theoretical framework, we have zt = [it, itUS, rt, et, yt,

ytUS, mt, mt

US], Π and Γ1, Γ2,.., Γp-1 are (8×8) matrices of parameters, κ is a (8×1) vector of parameters and ut is a (8×1) vector of white noise errors. To determine the number of cointegrating vectors in zt, we look at both maximum eigenvalue and trace tests. In case they don’t lead to the same conclusion, we rely on the maximum eigenvalue test as, in comparison, the trace test may lack power (Johansen and Juselius, 1990). We test for possible instability in the long-term relations using a stability analysis of the recursive eigenvalues (Hansen and Johansen, 1999). As in Barassi et al. (2005), we estimate time-varying adjustment coefficients using unobserved components models. As the different specification tests for α—AR(1), random walk with or without drift—result unsuccessful, we conclude that coefficients α are stable and that there is no structural breaks in the causal linkages that generate the cointegrating relations between the series. We check and validate the hypotheses on residuals, namely, no-serial correlation with the Ljung-Box statistic and normality of the distribution with the Jarque-Bera statistic.

To investigate more precisely the relations between the variables of each country’s VECM, we use an impulse response analysis. The impulse response function traces the effect of a one-time shock to one of the innovations on current and future values of the endogenous variables. However if the components of ut are contemporaneously correlated, and hence Σu is not diagonal, the shocks are not likely to occur in isolation in practice and impulse responses may not reflect the actual reactions of a given system properly. As explained by Lütkepohl (2007), orthogonalized shocks are often considered in impulse response analysis.

1 Augmented Dickey-Fuller (Dickey and Fuller, 1981).2 Kwiatkowski, Phillips, Schmidt and Shin (1992).


Any nonsingular matrix P with the property that PP’ = Σu can be used to define orthogonalized shocks as ηt = P-1ut where ηt are the structural shocks of the model and are contemporaneously uncorrelated. However if the residuals correlation matrix is diagonal, then the reduced-form model is identical to its structural form. As the matrix P is not unique, we have to impose restrictions on P, based on economic theory, which result in unique impulse responses. A popular choice of P is a lower triangular matrix obtained by a Choleski decomposition of Σu. For each country studied, we start by looking at the residuals correlation matrix and in case it is not diagonal, we check whether an alternative ordering of the variables would lead to modified responses. Impulse response functions are displayed as graphs. We consider one-standard-deviation shocks and look at the effects of the shocks on a 2� months period. Each impulse response goes with a 95% confidence band.

�. Empirical Results

The results of the Johansen cointegration tests as well as the set of linear restrictions on both α and β coefficients are presented in the Appendix, Table A3. The weak exogeneity tests enable us to determine, for each country, whether the domestic interest rate is the dependent variable in one of the cointegrating vectors3. We have introduced dummy variables where necessary to account for outliers. These control variables are detailed in the Appendix, Table A�. We do find cointegrating relations with the domestic interest rate being driven by some of the system variables in all eight countries. We concentrate on these interest rate equations. In the following section we present the long and short-run dynamics of the hard-peg countries’ domestic interest rates while the other one discusses those of control countries.

�.1 Hard-peg countries

Before proceeding to a detailed description of the econometric results, we include in Table 1, a summary of the results for the hard-peg countries. This table presents the long and short-run determinants of domestic interest rates as well as the impulse variables that prove to induce significant interest rate responses.

In all three countries, we find out that not only US variables but also internal variables play a role in the interest rate determination in the long and short run. However when analyzing impulse responses, it appears that the domestic interest rates of both dollarized countries are only influenced by US variables, namely the US interest rate and the US money supply. On the contrary, in the case of Argentina restricted to the currency board period, the only impulse variable that has an impact on local interest rates is the domestic money supply. We are tempted to interpret

3 Lag order selection criteria, recursive-eigenvalue stability tests and residual tests are not presented for a matter of space but are available upon request.

98 Cuadernos de Economía Vol. �7 (Mayo) 2010TA

BL

E 1

OV

ER

VIE

W O

F R

ESU

LTS

Det

erm

inan

ts o

f H

ard-

peg

Cou

ntri

es I

nter

est R

ates

Lon

g-ru

n de

term

inan

tsSh

ort-

run

dete

rmin

ants

Impu

lse

vari

able

s ha

ving

an

impa

cton

loca

l int

eres

t rat

es

Pana

ma

(199

8:2

– 20

07:8

)i tU

S , y

t, mtU

S∆

∆∆

iy

mt

tUS

tUS

−−

−1

22

,,

i tUS ,

mtU

S

Ecu

ador

(200

0:5

– 20

09:2

)i tU

S , y

tUS ,

mt, m

tUS

∆r t−

2i tU

S , m

tUS

Arg

entin

a(1

998:

2 –

2001

:9)

y tUS ,

mt

∆i t−

1, ∆

r t−2

mt

Sour

ce: A

utho

r el

abor

atio

n.N

otes

: itU

S : U

S in

tere

st ra

te, y

t: Dom

estic

leve

l of i

ncom

e, m

tUS :

US

mon

ey s

uppl

y, i t: D

omes

tic in

tere

st ra

te, y

US :

US

leve

l of i

ncom

e, m

t: Dom

estic

mon

ey s

uppl

y, r

: Ris

k pr

emiu

m.

TAB

LE

2T

HE

LO

NG

-RU

N D

ET

ER

MIN

AN

TS

OF

DO

ME

STIC

IN

TE

RE

ST R

AT

ES

i t

Cou

ntry

Con

stan

ti tU

Sr t

e ty t

y tUS

mt

mtU

Str

end

Pana

ma

1.6�

0.35

0N

/A0.

110

0-0

.31

(199

8:2

– 20

07:8

)(6

.66)

(2.7

8)(3

.37)

(-5.

�8)

Ecu

ador

-2.8

90.

�80

N/A

0-0

.32

-0.0

7-0

.1�

0.00

2(2

000:

5 –

2009

:2)

(2.9

0)(-

2.15

)(-

3.80

)(-

2.51

)(3

.52)

Arg

entin

a ha

rd p

eg2.

990

0N

/A0

-0.1

9-0

.20

0(1

998:

2 –

2001

:9)

(5.3

7)(-

3.57

)(-

6.17

)

Sour

ce: O

wn

estim

atio

n.N

otes

:Val

ues

in p

aren

thes

es a

re t-

stat

istic

s. i tU

S : U

S in

tere

st r

ate,

yt: D

omes

tic le

vel o

f in

com

e, m

tUS :

US

mon

ey s

uppl

y, i t: D

omes

tic in

tere

st r

ate,

yU

S : U

S le

vel o

f in

com

e,

mt: D

omes

tic m

oney

sup

ply,

r: R

isk

prem

ium

, et:

Exc

hang

e ra

te.


this as being due to the difference between a dollarization exchange rate regime and a currency-board regime, however results for Argentina have to be taken with caution as the sample is very short. Hard-peg countries’ interest rates are mostly influenced by US variables however not exclusively!

We below proceed to a detailed analysis of the econometric results. The long-run determinants of domestic interest rates in Panama as well as in Ecuador and Argentina restricted to their hard-peg period are presented in Table 2.

The long-run equations are all stable during the observation period. As expected, in the long run, we observe a positive impact of US interest rates on both dollarized countries’ interest rates, but we also notice the influence of domestic fundamentals! The monetary policy of both countries is not solely caused by US variables but is also oriented towards internal goals. We observe a positive influence of the domestic level of activity in the long-run equation of Panama’s interest rates. An increase in income raises the demand for money compared to the supply, generating an increase in the nominal interest rate. We also notice a negative impact of the domestic money supply on Ecuador and Argentina’s� interest rates. When there is a contraction of money supply relative to money demand, without a matching fall in prices, the domestic interest rate rises. In terms of foreign influence, we observe a negative impact of the US level of income on both countries’ rates, as expected theoretically. Finally, there is a negative influence of the US money supply on Panamanian and Ecuadorian interest rates, which is opposite to the sign given by the conceptual framework and may be due to full dollarization in both countries. Our interpretation of this result is that a rise in US money supply leads to a decline in the US interest rate which directly spills-over to both dollarized countries’ domestic interest rates. The case of Argentina during the currency board has to be interpreted with caution as the sample is very short. Table 3 presents the determinants of domestic interest rate changes.

� Restricted to the currency-board period.

TABLE 3THE SHORT-RUN DETERMINANTS

OF THE DOMESTIC INTEREST RATE DYNAMICS ∆it

Panama ∆ ∆ ∆ ∆i i y mt t tUS

tUS= + + −− − −0 28 0 05 0 05 01

� 332

2 102

2 39. . . .

( . ) ( . ) ( . )003 1

3 7�εt−

−( . )

Ecuador ∆ ∆it t t= + −−

− −−

0 001 0 02 0 1�2 05

23 09

16 56

. . .( . ) ( . ) ( . )

r ε

Argentina(hard peg)

∆ ∆ ∆i it t t t= − −− −−

−−

0 72 0 �6 0 8�13 12

22 10

16 80

. . .( . ) ( . ) ( . )

r ε

Source: Own estimation.Notes: Values in parentheses are t-statistics; εt captures the errors of the cointegrating relationship of Table 2.


Hard-peg-countries’ interest rate changes are determined by changes in both US and internal variables as well as by an error correction term. Changes in the risk premium are statistically significant in explaining changes in both Ecuadorian and Argentinian5 interest rates. The VECM’s short-term dynamics show that domestic interest rates are equilibrium-correcting. The adjustment coefficients range from 3% in the case of Panama to 1�% for Ecuador and 8�% in the case of Argentina. We also compute half-life6 coefficients, namely, the required time for the interest rates to adjust back towards their equilibrium level by 50%. It takes 16 months for the deviation of Panamanian interest rates from their long-run value to fall by half, 2.6 months in the case of Ecuadorian interest rates, while it only takes 1.5 month for Argentinian rates. However this last case has to be interpreted with caution as the sample is substantially smaller.

Once observed the long and short-run determinants of hard-peg countries’ domestic interest rates, we model shocks on the variables intervening in the long-term relations to find out the ones having a significant impact on interest rates. Impulse response functions measure the effect of a one-standard-deviation change in one of the system variables on the domestic interest rate. Table � presents the residuals correlation matrices.

Looking at these matrices, we check whether the shocks are likely to occur in isolation in practice and whether impulse responses may reflect the actual reactions of the given system properly. In the case of Panama, correlations between residuals are rather small with the exception of the one between domestic and US interest rates, while in the case of Ecuador we have relatively strong correlations between the US level of income and the US money supply and also between domestic and US money supplies. We model impulse responses with an alternative ordering of the variables when the ordering has an impact on responses. Finally, in the case of Argentina there is a 50% residuals correlation between the domestic interest rate and the domestic money supply and an also substantially high correlation between the domestic money supply and the US level of income. Figures 1 to 3 present the effect of successive shocks on the domestic interest rates in Panama, Ecuador and Argentina. The impulse responses are presented with a 95% confidence band.

In Panama, as presented in Figure 1, we first model a restrictive US monetary policy shock. It leads to a positive and significant response of the Panamanian interest rate on the entire period while a domestic demand shock has no effect at all. Finally a positive shock on the US money supply leads to a positive effect on the domestic interest rate, significant on the 2� months, which is in line with the conceptual framework but is opposite to the long-term relation we have between the local interest rate and the US money supply. However when we look at the short-run dynamics of the Panamanian interest rate, we effectively have a positive relation between changes in the

5 Restricted to the currency-board period.6 The half-life coefficient is defined as HL = ln(0.5)/ln(m) with εt = mεt−1 + Σj

k=1 τj−1∆εt−j + ηt

(Rossi, 2002).


TAB

LE

�R

ESI

DU

AL

S C

OR

RE

LA

TIO

N M

AT

RIC

ES

Pana

ma

Ecu

ador

Arg

entin

a

i ti tU

Sy t

mtU

Si t

i tUS

y tUS

mt

mtU

Si t

y tUS

mt

i t1

0.22

70

0.03

61

0.10

0.07

50.

100.

0�1

10.

007

0.50

i tUS

0. 2

271

0.08

30.

016

0.10

10.

105

0.02

�0.

056

y t0

0.08

31

0.01

8

y tUS

0.07

50.

105

10.

053

0.30

0.00

71

0.22

mt

0.10

0.02

�0.

053

10.

200.

500.

221

mtU

S0.

036

0.01

60.

018

10.

0�1

0.05

60.

300.

201

Sour

ce: O

wn

estim

atio

n.N

otes

: itU

S : U

S in

tere

st r

ate,

yt: D

omes

tic le

vel o

f in

com

e, m

tUS :

US

mon

ey s

uppl

y, i t: D

omes

tic in

tere

st r

ate,

yU

S : U

S le

vel o

f in

com

e, m

t: Dom

estic

mon

ey s

uppl

y. D

ata

from

Arg

entin

a is

res

tric

ted

to th

e cu

rren

cy b

oard

per

iod.


FIGURE 1PANAMANIAN INTEREST RATE RESPONSES

TO A SHOCK ON US INTEREST RATE, ON DOMESTIC INCOMEAND ON US MONEY SUPPLY

Source: Own estimation.


domestic interest rate and changes in the US money supply. The modeling of an alternative variables ordering has only a minor impact on the effect of the US money supply shock on domestic interest rates. Indeed the impact stays positive and significant, however just on the first part of the observation period and not any more on the 2� months.

FIGURE 2ECUADORIAN INTEREST RATE RESPONSES

TO A SHOCK ON US INTEREST RATE, ON US INCOME AND ON DOMESTIC AND US MONEY SUPPLIES


In the case of Ecuador, as reported in Figure 2, we also first model a restrictive US monetary policy shock that leads to a positive response of the domestic interest rate, significant as of the fourth month and on the rest of the period, while an external demand shock and a domestic monetary shock have almost no impact at all on Ecuadorian interest rates. Finally, as in Panama, a positive shock on the US money supply leads to a positive and significant effect on the domestic interest rate, in line with the conceptual framework but opposite to the long-run relation estimated. When we model an alternative ordering of the variables, we obtain domestic interest rate responses substantially different. As presented in Figure A1, in Appendix, an expansive


domestic monetary shock, when happening first, leads to a small increase of the domestic interest rate, a US monetary shock has almost no effect, while a positive external demand shock, with this alternative ordering of the variables, has a positive and significant impact on the Ecuadorian interest rate.

Finally, Figure 3 traces the effects of an external demand shock and a domestic monetary shock on the Argentinian interest rate, on the hard-peg period. The external demand shock has almost no significant impact on the domestic interest rate while, as expected theoretically, an expansive domestic monetary shock leads to a decline of the Argentinian interest rate, significant on the 2� months. The modeling of an alternative ordering of the variables doesn’t make any difference on the domestic interest rate responses.

FIGURE 3ARGENTINIAN INTEREST RATE RESPONSES TO A SHOCK

ON US INCOME AND ON DOMESTIC MONEY SUPPLY


The presence of domestic income in the long-run equation of Panamanian interest rates, of domestic money supply in the long-run equation of Ecuadorian interest rates as well as the presence of the risk premium in the latter’ interest rates short-run dynamics prove that both dollarized countries’ interest rates are not solely determined by US variables. However when analyzing impulse responses, it appears that domestic interest rates are exclusively influenced by US variables, namely the US interest rate and the US money supply in both dollarized countries. In the two countries, the fact that internal variables play a role in the interest rate determination in the long and short run make us conclude that their interest rates, even if mostly influenced by US variables as shown in the impulse responses analysis, are not exclusively caused by them! Argentina’s interest rate equilibrium relation includes the US level of income as well as the domestic money supply. However when simulating shocks on both variables, it appears that the domestic interest rate only respond to a shock on the domestic variable. We are tempted to interpret the difference between these three countries as being caused by the


exchange rate regime, namely dollarization regime and currency-board regime, however results for Argentina have to be taken with caution as the sample is very short. In the next section, we look at the behavior of short-term interest rates in the floating-exchange-rate countries.

�.2 Control countries

As in the previous section, we include in Table 5, an overview of the results for the control countries, before proceeding to a detailed description of the econometric results. With the exception of Brazil, we observe an impact of both domestic and US variables in the determination of the control-countries’ domestic interest rates looking at the long and short-run dynamics as well as looking at the impulse response functions. Domestic interest rates are not independent from what is happening in the United States as they are caused by both internal and US factors.

We below proceed to a detailed description of the econometric results of the control-countries’ domestic interest rates. The long-run determinants of the local interest rates of Brazil, Colombia, Mexico, Peru, Venezuela and Argentina restricted to the floating period, are presented in Table 6. The equilibrium equations are all stable on the observation period with the exception of Brazil at the end of 1998, beginning of 1999, corresponding to the deep financial crisis the country went through. We observe a foreign influence on local interest rates in the long run meaning that the monetary policy of control countries does not seem to exclusively pursue domestic aims. Mexican, Peruvian and Venezuelan rates all positively depend on US interest rates. There is also a positive influence of the exchange rate on Venezuelan and Argentinian rates, which turns negative in the case of Colombian rates. According to our theoretical framework, an exchange rate depreciation causes a rise in the domestic nominal interest rate as stated in the uncovered interest parity condition. However, according to the currency-substitution phenomenon, an exchange rate depreciation generates a fall in money demand relative to money supply, leading to a temporary decline in the domestic interest rate.

In terms of foreign influence, we also observe a negative impact of the US level of activity on Peruvian and Venezuelan rates and a positive influence of US money supply on Mexican rates. An expansion in the foreign money supply implies a depreciation of the exchange rate and then supposedly an increase of the domestic nominal interest rate. We also notice in each cointegrating equation the presence of internal factors. Namely, we observe a positive influence of the domestic level of activity for Brazilian, Mexican and Argentinian rates, as an increase in income raises the demand for money that generates an increase in the nominal interest rate. There is a negative influence of the domestic money supply for Brazilian, Colombian, Mexican and Argentinian rates as a contraction of money supply relative to money demand raises the domestic interest rate. We only model Brazilian rates after the break date, as of June 1999. This last case stands apart as we don’t get any direct foreign influence in the long run on


TAB

LE

5

OV

ER

VIE

W O

F R

ESU

LTS,

TH

E D

ET

ER

MIN

AN

TS

OF

CO

NT

RO

L-C

OU

NT

RIE

S IN

TE

RE

ST R

AT

ES

Lon

g-ru

n de

term

inan

tsSh

ort-

run

dete

rmin

ants

Impu

lse

vari

able

s ha

ving

an

impa

cton

loca

l int

eres

t rat

es

Bra

zil

(199

9:6

– 20

09:2

)r t, y

t, mt

∆i t−

1, ∆

i t−2,

∆r t−

1, ∆

r t−2,

∆

e t−1,

∆y t−

1, ∆

y t−2

r t, yt, m

t

Col

ombi

a(1

999:

8 –

2009

:3)

e t, mt

∆i t−

1, ∆

r t−1,

∆m

t−1

e t, mt

Mex

ico

(199

8:2

– 20

09:2

) i tU

S , y

t, mt,

mtU

S∆

i t−2,

∆i t−

2, ∆

r t−1,

∆y tU

S −1

y t, mtU

S

Peru

(199

8:2

– 20

09:3

)i tU

S , y

tUS

∆i t−

1, ∆

r t−1,

∆e t−

2, ∆

mt−

1i tU

S

Ven

ezue

la(1

998:

2 –

2009

:3)

i tUS ,

et,

y tUS

∆i t−

2, ∆

iUS t−

1, ∆

e t−1,

∆y t−

1i tU

S , e

t

Arg

entin

a(2

003:

8 –

2009

:4)

e t, yt,

mt, m

tUS

∆i t−

1, ∆

iUS t−

1, ∆

e t−1,

∆y t−

1e t, y

t, m

t

Sour

ce: O

wn

estim

atio

n.N

otes

: Val

ues

in p

aren

thes

es a

re t-

stat

istic

s. i tU

S : U

S in

tere

st ra

te, y

t: Dom

estic

leve

l of i

ncom

e, m

tUS :

US

mon

ey s

uppl

y, i t: D

omes

tic in

tere

st ra

te, y

US :

US

leve

l of i

ncom

e,

mt: D

omes

tic m

oney

sup

ply,

r: R

isk

prem

ium

, et:

Exc

hang

e ra

te.


TAB

LE

6

TH

E L

ON

G-R

UN

DE

TE

RM

INA

NT

S O

F D

OM

EST

IC I

NT

ER

EST

RA

TE

S i t

Cou

ntry

Con

stan

ti tU

Sr t

e ty t

y tUS

mt

mtU

S

Bra

zil

-3.3

8−

2.29

−1.

�2−

-0.2

7−

(199

9:6

– 20

09:2

)(1

3.10

)(6

.29)

(-5.

30)

Col

ombi

a1.

10−

−-0

.07

−−

-0.0

�−

(199

9:8

– 20

09:3

)(6

.75)

(-�.

0�)

(-7.

01)

Mex

ico

-12.

�52.

01−

−1.

01−

-0.9

12.

81(1

998:

2 –

2009

:2)

(-�.

63)

(2.1

1)(1

.99)

(-6.

07)

(6.7

3)

Peru

3.37

2.68

−−

−-0

.76

−−

(199

8:2

– 20

09:3

)(�

.96)

(7.0

1)(-

5.07

)

Ven

ezue

la6.

113.

05−

0.12

−-1

.3�

−−

(199

8:2

– 20

09:3

)(2

.�6)

(2.0

0)(-

2.21

)

Arg

entin

a−

−0.

�61.

32−

-0.�

�-0

.25

(200

3:8

– 20

09:4

)(6

.6�)

(11.

96)

(-10

.25)

(-3.

1�)

Sour

ce: O

wn

estim

atio

n.N

otes

: Val

ues

in p

aren

thes

es a

re t-

stat

istic

s. i tU

S : U

S in

tere

st ra

te, y

t: Dom

estic

leve

l of i

ncom

e, m

tUS :

US

mon

ey s

uppl

y, i t: D

omes

tic in

tere

st ra

te, y

US :

US

leve

l of i

ncom

e,

mt: D

omes

tic m

oney

sup

ply,

r: R

isk

prem

ium

, et:

Exc

hang

e ra

te.


domestic interest rates. They positively depend on the risk premium. A rise in country risk implies a rise in domestic interest rates as investors need to get a higher return for bearing the risk. Brazilian rates are also positively influenced by the domestic level of activity and negatively by the domestic money supply. Table 7 presents the determinants of domestic interest rate changes.

As indicated by the estimated error-correction models, in the short run, changes in both US and domestic variables are statistically significant in explaining changes in the control-countries’ rates. As in the case of hard-peg countries, we observe the impact of changes in the risk premium on the domestic interest rates for most countries, namely, Brazil, Mexico, Colombia and Peru. The adjustment coefficient is only 1% in the case of Peru, 5% for Brazil, Mexico and Colombia and is as high as 31% for floating Argentina and, finally, 36% for Venezuela. It takes more than 10 years for Peruvian interest rates to revert back to half the distance of their deviation to the long-run value while this same required time is a year and a half for Colombian interest rates and 7 months for Mexican rates. Half-lives are much smaller for the remaining three countries, 3.8 months in the case of Brazil, 3.5 for Venezuela and, lastly, only slightly more than a month for Argentina. We can’t really draw any conclusion from the computation of half-life coefficients of hard-peg countries, on one hand, control countries on the other hand, regarding any possible larger temporary autonomy for the ones or the others.

Overall, within these eight countries, Brazil appears to be the only one to have true autonomy as its interest rate is only caused by internal factors. In all other countries, whatever the exchange rate regime, both domestic and foreign variables determine the local interest rates in the long and short run. To find out which variables have a significant impact on interest rates, we model shocks on the variables intervening in the long-term relations. Impulse response functions measure the effect of a one-standard-deviation change on the response variable, namely the domestic interest rates. Table 8 presents the residuals correlation matrices.

In the case of Argentina restricted to the floating period, there is strong residuals correlation between the domestic interest rate and the exchange rate, between the domestic level of activity and the exchange rate and, finally, between the US money supply and the exchange rate. We model impulse responses with an alternative ordering of the variables when the ordering has an impact on responses, which is the case of Argentina. As far as Brazil is concerned, there is a relatively strong residuals correlation between the domestic money supply and the risk premium, while for Colombia and Mexico, correlation is high between the domestic interest rate and the domestic money supply. Finally, in the cases of both Peru and Venezuela, residuals correlations are quite small, meaning that shocks are likely to occur in isolation.

Figures � to 9 trace the current and future effects of a one-standard-deviation change in each variable intervening in the long-run relation on the domestic interest rates in all six countries. The impulse responses are presented with a 95% confidence band. With the exception of Argentina, the modeling of alternative variables orderings has no significant impact on the responses of domestic interest rates.


TAB

LE

7

TH

E S

HO

RT-

RU

N D

ET

ER

MIN

AN

TS

OF

TH

E D

OM

EST

IC I

NT

ER

EST

RA

TE

DY

NA

MIC

S ∆

i t

Bra

zil

∆i t =

0.6

8∆i t−

1 −

0.2

0∆i t−

2 −

0.1

0∆r t−

1 −

0.0

8∆r t−

2 −

0.0

2∆e t−

1 −

0.0

9∆y t−

1 −

0.0

5∆y t−

2 −

0.0

5εt−

1

(12.

09)

(−�.

07)

(−3.

89)

(−3.

33)

(−2.

72)

(−5.

27)

(−2.

65)

(−9.

9�)

Col

ombi

a∆

i t = 0

.7�∆

i t−1

+ 0

.11∆

r t−1

− 0

.07∆

mt−

1 −

0.0

5εt−

1

(11.

63)

(2.�

1)

(−5.

0�)

(−�.

3�)

Mex

ico

∆∆

∆∆

∆i

iy

tt

it

=−

++

−−

−−

033

281

180

0��

22

�22

298

13

65.

..

.(

.)

(.

)(

.)r

ttUS

t−

−−

− 12

301

272

005

(.

)(

.)

.ε

Peru

∆i t =

0.5

8∆i t−

1 +

0.0

8∆r t−

1 +

0.0

3∆e t−

2 +

0.0

2∆m

t−1

− 0

.01ε

t−1

(6

.�7)

(3

.15)

(2

.76)

(2

.33)

(−

2.71

)

Ven

ezue

la∆

i t = 0

.17∆

i t−2

− 5

.68∆

i tUS −1

+ 0

.28∆

e t−1

− 0

.1�∆

y t−1

− 0

.37ε

t−1

(2

.20)

(−

2.�6

) (3

.51)

(−

3.06

) (−

7.61

)

Arg

entin

a∆

i t = 0

.�9∆

i t−1

+ 1

.20∆

iUS t−

1 +

0.1

8∆e t−

1 −

0.1

9∆y t−

1 −

0.3

1εt−

1

(�.9

�)

(2.5

1)

(3.1

5)

(−2.

51)

(−5.

75)

Sour

ce: O

wn

estim

atio

n.N

otes

: Val

ues

in p

aren

thes

es a

re t-

stat

istic

s, ε

t cap

ture

s th

e er

rors

of

the

coin

tegr

atin

g re

latio

nshi

p of

Tab

le 6

.


BL

E 8

R

ESI

DU

AL

S C

OR

RE

LA

TIO

N M

AT

RIC

ES

Arg

entin

a (f

loat

)B

razi

lC

olom

bia

i te t

y tm

tm

tUS

i tr t

y tm

ti t

e tm

t

i t1

0.30

50.

008

0.02

0.19

10.

008

0.10

80.

201

-0.2

05-0

.32

e t0.

305

10.

270.

080.

28-0

.205

10.

15

r t0.

008

10.

092

0.29

y t0.

008

0.27

10.

03�

0.08

0.10

80.

092

10.

068

mt

0.02

0.08

0.03

�1

0.11

0.20

0.29

0.06

81

-0.3

20.

151

mtU

S0.

190.

280.

080.

111

Mex

ico

Peru

Ven

ezue

la

i ti tU

Sy t

mt

mtU

Si t

i tUS

y tUS

i ti tU

Se t

y tUS

i t1

0.1�

-0.0

�7-0

.2�

0.21

10.

055

0.09

�1

-0.0

86-0

.02

0.01

7

i tUS

0.1�

10.

10-0

.025

-0.0

820.

055

10.

0�8

-0.0

861

0.06

30.

0�

e t-0

.02

0.06

31

0.15

5

y t-0

.0�7

0.10

10.

057

-0.0

95

y tUS

0.09

�0.

0�8

10.

017

0.0�

0.15

51

mt

-0.2

�-0

.025

0.05

71

0.09

3

mtU

S0.

21-0

.082

-0.0

950.

093

1

Sour

ce: O

wn

estim

atio

n.N

otes

: Val

ues

in p

aren

thes

es a

re t-

stat

istic

s. i tU

S : U

S in

tere

st ra

te, y

t: Dom

estic

leve

l of i

ncom

e, m

tUS :

US

mon

ey s

uppl

y, i t: D

omes

tic in

tere

st ra

te, y

US :

US

leve

l of i

ncom

e,

mt: D

omes

tic m

oney

sup

ply,

r: R

isk

prem

ium

, et:

Exc

hang

e ra

te.


FIGURE �BRAZILIAN INTEREST RATE RESPONSES TO A SHOCK

ON RISK PREMIUM, ON DOMESTIC INCOMEAND ON DOMESTIC MONEY SUPPLY



In the case of Brazil, as shown in Figure �, we first model a negative shock on the country’s investor confidence. It leads to a significant increase of the domestic interest rate. The response is significant after 6 months and during a year and is significant again after 21 months. We then model a positive domestic demand shock. It leads to a significant increase of the interest rate on the entire period. Finally, we model an expansive domestic monetary shock and find out that it has only a minor effect on the domestic interest rate.

FIGURE 5COLOMBIAN INTEREST RATE RESPONSES TO A SHOCK

ON EXCHANGE RATE AND ON DOMESTIC MONEY SUPPLY


When we model impulse response functions for Colombia, presented in Figure 5, we find out that a depreciation leads to a decline of the interest rate as stated in the long-run relation. According to the currency-substitution phenomenon, an exchange rate depreciation generates a fall in money demand relative to money supply, leading to a temporary decline in the domestic interest rate. This decline is significant after 9 months while an expansive domestic monetary shock leads to an immediate fall of the domestic interest rate, significant on a 1� months period.

Figure 6 presents the case of Mexico. A restrictive US monetary policy shock and an expansive domestic monetary shock have no impact on the domestic interest rate whatever the ordering of the variables. A positive domestic demand shock induces an increase of the interest rate significant after 9 months and on the rest of the period while a positive shock on the US money supply leads to an increase in the interest rate, as stated in the long-run relation estimated as well as in the conceptual framework. This increase is significant after 10 months and on the rest of the period.

Figure 7 traces the effects of a US monetary policy shock and a US demand shock on the Peruvian interest rate. The restrictive US monetary shock leads to an increase of the Peruvian interest rate, significant on the second part of the observation period, while the shock in the external demand has no impact at all.


FIGURE 6MEXICAN INTEREST RATE RESPONSES TO A SHOCK

ON US INTEREST RATE, ON DOMESTIC INCOMEAND ON DOMESTIC AND US MONEY SUPPLIES


FIGURE 7PERUVIAN INTEREST RATE RESPONSES TO A SHOCK

ON US INTEREST RATE AND ON US INCOME



FIGURE 8VENEZUELAN INTEREST RATE RESPONSES

TO A SHOCK ON US INTEREST RATE, ON EXCHANGE RATEAND ON US INCOME



In the case of Venezuela, presented in Figure 8, a restrictive US monetary policy shock leads to an immediate decline of the interest rate, which is opposite to the interest parity theory. Looking at the long-run determinants of the Venezuelan interest rate, we notice a positive relation with the US interest rate however, looking at the short-run determinants, we observe a negative impact of changes in the US interest rate. This impact is significant on the first 10 months of the observation period. A depreciation leads to an increase in the interest rate, as stated in the uncovered interest parity condition, however this increase shortly disappears. Finally, a positive US demand shock leads to a tiny decline of the domestic interest rate.

FIGURE 9ARGENTINIAN INTEREST RATE RESPONSES

TO A SHOCK ON EXCHANGE RATE, ON DOMESTIC INCOMEAND ON DOMESTIC AND US MONEY SUPPLIES


Finally, on Figure 9, we look at the behavior of Argentinian interest rates on the country’s floating rates period. A depreciation of the exchange rate leads to a tiny increase in the interest rate which disappears after � months. As by our conceptual framework, a positive shock on the domestic demand leads to an increase in the domestic interest rate, immediate and significant


on the entire period, while an expansive domestic monetary shock induces an immediate and significant decline of the interest rate. Finally, an expansive US monetary shock has no impact on the interest rate. We present in the Appendix, Figure A2, an alternative variables ordering as responses are then slightly different. Instead of a shock on the exchange rate happening first, we model the exchange rate disturbance as being the last shock. The interest rate response is then larger and significant on the 2� months period. As in the case of Venezuela, a depreciation leads to an increase in the interest rate, as stated in the uncovered interest parity condition. As far as the three other variables are concerned, we observe smaller interest rate responses following a positive domestic demand shock and following an expansive domestic monetary shock, however responses are still significant. Whatever the variables ordering, the US monetary shock has no impact at all.

With the exception of Brazil, we observe an impact of both domestic and US factors in the determination of control-countries’ domestic interest rates looking at the long and short-run dynamics as well as looking at the impulse response functions. Domestic interest rates are not independent from what is happening in the United States. They are caused by both domestic and US variables. We conclude that the monetary policy of control countries does not exclusively pursue domestic aims, with the exception of the region’s largest country, Brazil, the only floating-rate-economy of our sample that proves to benefit from monetary autonomy.

5. Conclusions

We find empirical evidence that in Latin America the exchange rate regime does not rigidly determine the degree of monetary policy independence. Hard-peg countries enjoy some independence as their interest rates are mostly but not exclusively determined by US variables while, even perfectly-flexible rates may not guarantee monetary autonomy since the interest rates of our set of control-countries, with the exception of Brazil, are not only determined by internal factors but also by US variables. Unlike the traditional view, we conclude that economies with rigidly-fixed exchange rates do not bear a loss of monetary autonomy substantially higher than that of floating-exchange-rate economies. When currency substitution exists, even perfectly flexible rates may not guarantee monetary independence. As stated in the “fear of floating” literature, countries that are not pegging their currencies do often choose to follow the base interest rate to some degree, not because they cannot exert independence but because they simply choose not to do so. The potential instability of floating rates does not seem to be effectively compensated by any meaningful monetary freedom. In terms of policy implications and as explained by Canova (2005), it seems that Latin American policymakers are required to carefully monitor US conditions, whatever the exchange rate regime!


REFERENCES

Adam, C., D. Cobham, and E. Girardin (2005), “Monetary Frameworks and Institutional Constraints: UK Monetary Policy Reaction Functions, 1985-2003”, Oxford Bulletin of Economics and Statistics, 67: �97-516.

Alper, E., O. Ardic, and S. Fendoglu (2007), “The Economics of Uncovered Interest Parity Condition for Emerging Markets: a Survey”, Journal of Economic Surveys, 23(1): 115-138.

Arize, A. (199�), “A Re-examination of the Demand for Money in Small Developing Economies,” Applied Economics, 26(3):217-28.

Barassi, M., G. Caporale, and S. Hall (2005), “Interest Rate Linkages: a Kalman Filter Approach to Detecting Structural Change”, Economic Modeling, 22(2): 253-28�.

Borensztein, E., J. Zettelmeyer, and T. Philippon (2001), “Monetary Independence in Emerging Markets: Does the Exchange-Rate Regime Make a Difference?”, IMF Working Paper 01/1.

Calvo, G. and C. Reinhart (2002), “Fear of Floating”. Quarterly Journal of Economics, 117(2): 379-�08.

Canova, F. (2005), “The Transmission of US Shocks to Latin America”, Journal of Applied Econometrics, 20(2): 229-251.

Clarida, R., J. Galí, and M. Gertler (1998), “Monetary Policy Rules in Practice: some International Evidence”, European Economic Review, �2(6): 1033-1067.

Clarida R., J. Galí, and M. Gertler (1999), “The Science of Monetary Policy: a New Keynesian Perspective”, Journal of Economic Literature, 37: 1661-1707.

Clarida, R., J. Galí, and M. Gertler (2001), “Optimal Monetary Policy in Open versus Closed Economies: an Integrated Approach”, American Economic Review, 91(2): 2�8-252.

Dickey, D.A. and W.A. Fuller (1981), “Likelihood Ratio Statistics for Autoregressive Time Series with a Unit Root”, Econometrica, �9(�): 1057-1072.

Dornbusch, R. (1976), “Expectations and Exchange Rate Dynamics”, Journal of Political Economy, 8�(6): 1161-1176.

Frankel, J. (1979), “On the Mark: a Theory of Floating Exchange Rates based on Real Interest Differentials”, American Economic Review, 69(�): 610-622.

Frankel, J. (1999), “No Single Currency Regime is Right for all Countries or at all Times”, Working Paper 7338, NBER.

Frankel, J., S. Schmukler, and L. Servén (2002), “Global Transmission of Interest Rates: Monetary Independence and Currency Regime”, Working Paper 2�2�, World Bank Policy Research.

De Freitas, M. and F. Veiga (2002), “Currency Substitution, Portfolio Diversification, and Money Demand”, Canadian Journal of Economics, 39(3): 719-7�3.

Hansen, H. and S. Johansen (1999), “Some Tests for Parameter Constancy in Cointegrated VAR-models”, Econometrics Journal, 2: 306-333.

Hausmann R., M. Gavin., C. Pages Serra, and E. Stein (1999), “Financial Turmoil and the Choice of Exchange Rate Regimes”, Working Paper �00, Inter-American Development Bank.

Johansen, S. (1988), “Statistical Analysis of Cointegrating Vectors”, Journal of Economics, Dynamics and Control, 12(2): 231-25�.

Johansen, S. and K. Juselius (1990), “Maximum Likelihood Estimation and Inference on Cointegration, with Applications to the Demand for Money”, Oxford Bulletin of Economics and Statistics, 52: 169-210.


Kwiatkowski, D., P.C.B Phillips, P. Schmidt, and Y. Shin (1992), “Testing the Null Hypothesis of Stationarity against the Alternative of a Unit Root: How Sure are we that Economic Time Series Have a Unit Root?”, Journal of Econometrics, 5�(1-3): 159-178.

Levy-Yeyati, E. and F. Sturzenegger (2005), “Classifying Exchange Rate Regimes: Deeds vs. Words”, European Economic Review, �9(6): 1603-1635.

Lütkepohl, H. (2007), “Econometric Analysis with Vector Autoregressive Models”, Working Paper ECO2007/11, European University Institute Economics.

Miles, M. (1978), “Currency Substitution, Flexible Exchange Rates, and Monetary Independence”, American Economic Review, 68(3): �28-36.

Mundell, R. (1963), “Capital Mobility and Stabilization Policy under Fixed and Flexible Exchange Rates”, Canadian Journal of Economics and Political Science, 29: �75-85.

Ng, S. and P. Perron (2001), “Lag Length Selection and the Construction of Unit Root Tests with Good Size and Power”, Econometrica, 69: 1519-5�.

Obstfeld, M., J. Shambaugh, and A. Taylor (200�), “Monetary Sovereignty, Exchange Rates, and Capital Controls: the Trilemma in the Interwar Period”, IMF Staff Papers 51: 75-108.

Obstfeld, M., J. Shambaugh, and A. Taylor (2005), “The Trilemma in History: Tradeoffs among Exchange Rates, Monetary Policies, and Capital Mobility”, Review of Economics and Statistics, 87(3): �23-38.

Phillips, P.C.B. and P. Perron (1988), “Testing for a Unit Root in Time Series Regression”, Biometrika, 75(2): 335-3�6.

Rossi, B. (2002), “Confidence Intervals for Half-Life Deviations from Purchasing Power Parity”, Working Paper 02-08, Duke University, Department of Economics.

Shambaugh, J. (200�), “The Effect of Fixed Exchange Rates on Monetary Policy”, Quarterly Journal of Economics, 119(1): 300-51.


Appendix

TABLE A1DATA DESCRIPTION

Interestrate

Priceindex

Exchangerate

Monetaryaggregate

Riskpremium

Productionproxy

Panama Description 6 monthinterest rate

CPI N/A M1 SpreadEMBI+

IMAE (s.a.)

Sample 1998:12007:8

1998:12007:8

1998:12007:8

1998:12007:8

1998:12007:8

Source IFS IFS IFS JP Morgan CGRP

Ecuador Description Short-termdeposit rate


Crude petroleumproduction (s.a.)

Sample 2000:32009:2

2000:32009:2

2000:32009:2

2000:32009:2

2000:32009:2

Source IFS IFS IFS JP Morgan IFS

Argentina(hard peg)

Description Moneymarket rate


IMAE (s.a.)

Sample 1998:12001:9

1998:12001:9

1998:12001:9

1998:12001:9

1998:12001:9

Source IFS IFS IFS JP Morgan BCRA+

Argentina(float)

Description Moneymarket rate

CPI Official rate,end of period

M1 SpreadEMBI+

IMAE (s.a.)

Sample 2003:62009:�

2003:62009:�

2003:62009:�

2003:62009:�

2003:62009:�

2003:62009:�

Source IFS IFS IFS IFS JP Morgan BCRA

Brazil Description Moneymarket rate

CPI Market rate,end of period

M1 SpreadEMBI+

Industrialproduction (s.a.)

Sample 1998:12009:2

1998:12009:2

1998:12009:2

1998:12009:2

1998:12009:2

1998:12009:2

Source IFS IFS IFS IFS JP Morgan IFS

Mexico Description Treasurybill rate

CPI Principal rate,end of period

M1 SpreadEMBI+

Industrialproduction (s.a.)

Sample 1998:12009:2

1998:12009:2

1998:12009:2

1998:12009:2

1998:12009:2

1998:12009:2


Colombia Description Moneymarket rate


M1 SpreadEMBI+

Manufacturingproduction (s.a.)

Sample 1999:62009:3

1999:62009:3

1999:62009:3

1999:62009:3

1999:62009:3

1999:62009:3


Peru Description Short-termdeposit rate

CPI Market rate,end of period

M1 SpreadEMBI+

Indice mensualde producción (s.a.)

Sample 1998:12009:3

1998:12009:3

1998:12009:3

1998:12009:3

1998:12009:3

1998:12009:3

Source IFS IFS IFS IFS JP Morgan BCRP

Venezuela Description Moneymarket rate


M1 SpreadEMBI+

Indice mensualde producción (s.a.)

Sample 1998:12009:3

1998:12009:3

1998:12009:3

1998:12009:3

1998:12009:3

1998:12009:3

Source IFS IFS IFS IFS JP Morgan BCV

United-States Description Treasurybill rate

CPI N/A M1 Industrialproduction (s.a.)

Sample 1998:12009:�

1998:12009:�

1998:12009:�

1998:12009:�

Source IFS IFS IFS IFS

Source: Own estimation.Notes: CPI: Consumer Prices Index; N/A: not applicable; IMAE: Índice Mensual de Actividad Económica; s.a.: seasonally adjusted; I.M.F. International Financial Statistics; CGRP: Contraloría General de la República de Panamá; BCRA: Banco Central de la República Argentina; BCRP: Banco Central de Reserva del Perú; BCV: Banco Central de Venezuela.


BL

E A

2U

NIT

RO

OT

TE

STS

RE

SULT

S

i te t

y tM

tr t

i te t

y tM

tr t

Pana

ma

AD

F I(

1)N

/AI(

1)I(

1)I(

1)E

cuad

orA

DF

I(1)

N/A

I(1)

I(1)

I(1)

PP

I(1)

N/A

I(1)

I(1)

I(1)

PP

I(1)

N/A

I(1)

I(1)

I(1)

KPS

S I(

1)N

/AI(

1)I(

1)I(

1)K

PSS

I(1)

N/A

I(1)

I(1)

I(1)

NP

I(1)

N/A

Inc

I(1)

I(0)

NP

I(1)

N/A

I(1)

I(1)

I(1)

Arg

entin

aA

DF

I(1)

I(1)

I(1)

I(1)

I(1)

Bra

zil

AD

F I(

1)I(

1)I(

1)I(

2)I(

1)PP

I(

1)I(

1)I(

1)I(

1)I(

1)PP

I(

1)I(

1)I(

1)I(

1)I(

1)K

PSS

I(1)

I(0)

I(1)

I(1)

I(0)

KPS

S I(

1)I(

1)I(

1)I(

1)I(

1)N

PI(

1)I(

1)In

cIn

cI(

1)N

PI(

1)I(

1)I(

1)I(

1)I(

1)

Col

ombi

aA

DF

I(1)

I(1)

I(1)

I(1)

I(2)

Mex

ico

AD

F I(

1)I(

1)I(

2)I(

2)I(

1)PP

I(

1)I(

1)I(

1)I(

1)I(

1)PP

I(

1)I(

1)I(

1)I(

1)I(

1)K

PSS

I(1)

I(1)

I(1)

I(1)

I(1)

KPS

S I(

1)I(

1)I(

1)I(

1)I(

1)N

PI(

1)I(

1)I(

1)I(

1)I(

1)N

PI(

1)I(

1)In

cI(

1)I(

1)

Peru

AD

F I(

1)I(

1)I(

2)I(

2)I(

1)V

enez

uela

AD

F I(

0)I(

1)I(

1)I(

1)I(

1)PP

I(

1)I(

1)I(

1)I(

1)I(

1)PP

I(

1)I(

1)I(

1)I(

1)I(

1)K

PSS

I(1)

I(1)

I(1)

I(1)

I(1)

KPS

S I(

1)I(

1)I(

1)I(

1)I(

0)N

PI(

1)I(

1)I(

1)I(

2)I(

1)N

PI(

1)I(

1)I(

1)I(

1)I(

1)

Uni

ted

Stat

es

AD

F I(

1)N

/AI(

2)I(

2)PP

I(

1)N

/AI(

1)I(

1)K

PSS

I(1)

N/A

I(1)

I(1)

NP

I(1)

N/A

I(1)

I(1)

Sour

ce: O

wn

estim

atio

n.N

otes

: i t: n

omin

al i

nter

est

rate

, e t:

Exc

hang

e ra

te,

y t: Rea

l in

com

e, M

t: M

onet

ary

aggr

egat

e, r

: Ris

k pr

emiu

m;

AD

F: A

ugm

ente

d D

icke

y-Fu

ller

. PP

: Phi

llip

s-Pe

rron

. K

PSS

: Kw

iatk

owsk

i, Ph

illip

s, S

chm

idt a

nd S

hin.

NP:

Ng-

Perr

on;

Inc.

: Inc

oncl

usiv

e. S

ampl

e fo

r Pa

nam

a is

the

sam

e of

Tab

le 1

.


TAB

LE

A3

CO

INT

EG

RA

TIO

N T

EST

S R

ESU

LTS

AN

D C

OE

FFIC

IEN

TS

RE

STR

ICT

ION

TE

STS

No.

of C

E

Tre

ndas

sum

ptio

nM

Est

atis

tic5%

cri

tical

va

lue

α a

nd β

coe

ffic

ient

s re

stri

ctio

n te

sts

LR

test

Pana

ma

r =

0r

= 1

r =

2

No

det.

tren

d 6

2.86

��.

37 3

2.03

�7.

08 �

0.96

3�.

81

β(2,

3)=

β(1,

1)=

1, β

(2,1

)=β(

2,2)

=β(

2,7)

=β(

2,5)

=β(

1,3)

=β(

1,5)

=β(

1,7)

=0,

α(1

,2)=

α(2

,2)=

α(5

,2)=

α(6

,2)=

α

(2,1

)=α

(7,1

)=α

(�,1

)=0

χ2 (

12)

= 2

0.53

[0.0

58]

Ecu

ador

r =

0r

= 1

Lin

ear

det.

tren

d 6

2.36

39.3

1 5

0.60

��.5

0β(

1,1)

=1,

β(1

,3)=

β(1,

�)=

0, α

(2,1

)=α

(3,1

)=α

(�,1

)= α

(5,1

)=α

(7,1

)=0

χ2 (

7) =

11.

02[0

.138

]

Arg

entin

aha

rd p

egr

= 0

r =

1N

o de

t. tr

end

66.

8� �

3.73

50.

60 �

�.50

β(1,

1)=

1, β

(1,2

)=β(

1,3)

=β(

1,�)

=β(

1,7)

=0,

α(7

,1)=

α(2

,1)

=α

(3,1

)=α

(�,1

)=α

(5,1

)=0

χ2 (

9) =

11.

78[0

.226

]

Arg

entin

afl

oat

r =

0r

= 1

r =

2

No

det.

tren

d 1

02.9

1 6

6.95

32.

79

53.

19 �

7.08

�0.

96

β(2,

7)=

β(1,

1)=

1, β

(1,

2)=

β(1,

3)=

β(1,

6)=

β(2,

1)=

β(2,

3)=

β(2,

8)=

0, (

6,1)

=α

(2,2

)=α

(�,2

)=α

(5,2

)=α

(6,2

)=α

(�,1

)= α

(5,1

)=α

(7,1

)=α

(8,1

)=α

(1,2

)=0

χ2 (

1�)

= 2

3.32

[0.0

55]

Bra

zil

r =

0r

= 1

r =

2

Lin

ear

det.

tren

d 9

8.�1

60.

00 3

9.98

52.

36 �

6.23

�0.

08

β(1,

1)=

β(2,

5)=

1, β

(2,2

)=β(

2,3)

=β(

1,8)

=β(

1,2)

=β(

1,6)

=β(

2,�)

=β(

2,7)

=β(

2,1)

=β(

1,�)

=0,

α(1

,2)=

α(8

,1)=

α(2

,1)=

α(3

,1)=

α(5

,1)=

α(6

,1)=

α(7

,1)=

α(6

,2)=

α(7

,2)=

0

χ2 (

16)

= 2

1.91

[0.1

�6]

Mex

ico

r =

0r

= 1

r =

2

No

det.

tren

d 8

0.5�

52.

00 3

9.�6

53.

19 �

7.08

�0.

96

β(1,

1)=

β(2,

7)=

1, β

(2,�

)=β(

2,3)

=β(

1,6)

=β(

2,2)

=β(

1,�)

=β(

1,3)

=β(

2,1)

=0,

α(8

,2)=

α(3

,1)=

α(�

,1)=

α(5

,1)=

α(6

,1)=

α(7

,1)=

α(1

,2)=

α(3

,2)=

α(�

,2)=

α(5

,2)=

α(6

,2)=

0

χ2 (

16)

= 2

5.08

[0.0

68]

Col

ombi

ar

= 0

r =

1N

o de

t. tr

end

75.

12 �

5.52

53.

19 �

7.08

β(1,

1)=

1, β

(1,8

)=β(

1,6)

=β(

1,2)

=β(

1,5)

=β(

1,3)

=0

α(6

,1)=

α(�

,1)=

α(8

,1)=

α(2

,1)=

α(3

,1)=

α(5

,1)=

0χ

2 (11

) =

13.

68[0

.251

]

Peru

r =

0r

= 1

r =

2

No

det.

tren

d 7

1.83

�9.

87 3

8.15

53.

19 �

7.08

�0.

96

β(1,

1)=

β(2,

3)=

1, β

(2,2

)=β(

2,6)

=β(

2,1)

=β(

2,5)

=β(

2,7)

=β(

1,�)

=β(

1,5)

=β(

1,7)

=β(

1,8)

=β(

1,3)

=0,

α(1

,2)=

α(5

,2)=

α(7

,2)=

α(8

,2)=

α(2

,2)=

α(6

,1)=

α(2

,1)=

α(3

,1)=

α(�

,1)=

0

χ2 (

17)

= 2

6.58

[0.0

6�]

Ven

ezue

lar

= 0

r =

1r

= 2

Lin

ear

det.

tren

d 7

2.17

50.

7�38

.33

52.

36 �

6.23

�0.0

8

β(1,

1)=

β(2,

5)=

1, β

(2,2

)=β(

2,1)

=β(

2,8)

=β(

2,6)

=β(

1,3)

=β(

1,5)

=β(

1,7)

=β(

1,8)

=0,

α(�

,2)=

α(3

,2)=

α(1

,2)=

α(2

,2)=

α(7

,2)=

α(8

,2)=

α(5

,1)=

α(7

,1)=

α(2

,1)=

α(�

,1)=

α(6

,1)=

α(8

,1)=

0

χ2 (

18)

= 2

7.62

[0.0

68]

Sour

ce: O

wn

estim

atio

n.N

otes

: N°.

CE

: Num

ber

of c

oint

egra

ting

equa

tion(

s); M

E: M

axim

um-E

igen

valu

e; L

R: L

ikel

ihoo

d R

atio

; det

.: det

erm

inis

tic.;

p-va

lues

in b

rack

ets.

Sam

ple

for

Pana

ma

is

the

sam

e of

Tab

le 2

.


FIGURE A1ECUADORIAN INTEREST RATE RESPONSES TO A SHOCK

ON DOMESTIC AND US MONEY SUPPLIES, ON US INTEREST RATE,AND ON US INCOME



FIGURE A2ARGENTINIAN INTEREST RATE RESPONSES TO A SHOCK

ON DOMESTIC INCOME, ON DOMESTIC AND US MONEY SUPPLIES, AND ON EXCHANGE RATE



TABLE A� INTERVENTION VARIABLES DETAILS

Argentina During the currency board period, three outliers, March, July and August 2001, coincide with the deep crisis and traumatic devaluation of the Peso the country went through.

Brazil A dummy is necessary in August 1999 as a consequence of the financial crisis and currency devaluation the country faced at the beginning of that same year. The outlier in February 2003 is supposedly linked with Argentina’s economic crisis.

Colombia Outliers in August, October and December 1999 can be attributed to the country’s move to a floating exchange rate regime in September 1999, after abandoning the crawling-peg band system introduced in 1992.

Ecuador There are outliers in September and October 2007 as president Correa is changing the country’s political landscape by rewriting the constitution. A dummy is necessary in November 2008, just before the president made good on months of threats by defaulting on a US$30 million coupon owed to Ecuador’s global ’12s, and on US$2.7 billion of global ’30s.

Mexico Outliers are present in 1999, February, April and October. We suppose they are a consequence of the Brazilian crisis.

Peru There is an outlier in September 2001, just before the decision of the Central Bank to explicitly target a range for CPI inflation. The outliers in March and June 2008 are linked to the measures announced by the Central Bank in April 2008, namely, an increase in the fee charged to foreigners on the purchase of “Certificates of Deposits” to �00 basis points and a hike in the marginal reserve requirement on PEN deposits by foreigners in local banks.

Venezuela We notice outliers in December 2001 and February 2002 as the country faces deep political troubles that have led to the “coup”, on the 11th of April, 2002. The dummy in December 2002 coincides with a 63-day strike the country faced. Strict capital controls have been in place since January 2003 as authorities have reacted to intense pressure on the currency and bank deposits generated by capital flight.

Source: Author elaboration.

Documents

The Impact of Exchange Rate Regime on Interest Rates in ... · Exchange Rate Regimes and Interest Rates in Latin America 93 (1979) model to take into account emerging countries’