Monetary policy and interest rates: evidence from Mexico

North American Journal of Economics and Finance 14 (2003) 357–379

Monetary policy and interest rates: evidence from Mexico Alberto Torres∗ Banco de México, Av. 5 de Mayo #18 5to piso, C.P. 06059, Mexico City, Mexico Received 7 April 2003; received in revised form 18 August 2003; accepted 21 August 2003

Abstract Using the recent experience of Mexico under a flexible exchange-rate regime and under an inflationtargeting framework, this paper examines the extent to which monetary policy has performed the role of nominal anchor for the Mexican economy. The paper identifies a set of variables together with a monetary policy rule, which offer a good approximation to the process through which interest rates are determined. The evidence suggests that recent monetary policy in Mexico has been consistent with inflation-targeting principles and has become the nominal anchor for the economy. © 2003 Elsevier Inc. All rights reserved. JEL classification: E52; F33 Keywords: Monetary policy rules; Inflation targeting; Flexible exchange rates

1. Introduction This paper analyzes the process through which interest rates are determined in Mexico. The purpose is to formally test whether under the current flexible exchange-rate regime and inflation-targeting framework, monetary policy has served as nominal anchor for the economy. Using the framework of monetary policy rules, this paper analyzes the process through which interest rates are determined in the Mexican economy. The analysis identifies a set of variables that, combined with a monetary policy rule, offer a good approximation to the process through which interest rates are determined. One of the key differences between a regime in which the exchange rate is fixed and one in which it floats, is the role of monetary policy in the economy. In the first case, monetary policy is constrained by the exchange rate and its role is to support that rate. The authorities ∗

Tel.: +52-55-5237-2729; fax: +52-55-5237-2571. E-mail address: [email protected] (A. Torres).

1062-9408/$ – see front matter © 2003 Elsevier Inc. All rights reserved. doi:10.1016/j.najef.2003.08.001

358

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

are responsible for price stability, and it is said that the exchange rate serves as the nominal anchor of the economy. Under a floating exchange-rate regime, on the other hand, monetary policy is not constrained by any rule and the monetary authorities are responsible for setting monetary policy in order to achieve price stability. In this case, monetary policy plays an important role in providing the nominal anchor of the economy. At the end of 1994, Mexico abandoned the fixed exchange-rate regime and adopted a flexible rate. The transition was anything but smooth, since the Mexican economy experienced a major financial crisis throughout 1995. As expected, the crisis impaired the credibility of Mexican financial and monetary institutions and thus made it more difficult for the Bank of Mexico to establish monetary policy as the nominal anchor of the economy. Over the years, however, monetary policy has evolved toward an inflation-targeting framework, allowing financial stability to be enhanced and inflation to be reduced. The recent Mexican experience is instructive for two reasons. First, it represents an emerging market economy under a flexible exchange-rate regime that has been able to reduce inflation from two-digit to one-digit levels. Therefore, if it can be shown that monetary policy has contributed to inflation reduction, then Mexico’s experience suggests that fixed-rate regimes, with endogenous monetary policy, are not necessarily the only choice for emerging market economies faced with high inflation. It suggests that an independent monetary policy can help reduce inflation. Second, in recent years, the instrumentation of monetary policy in Mexico has experienced major changes. After the 1995 crisis, instrumentation was focused primarily on intermediate targets such as domestic credit ceilings and a buildup of international reserves. Then, as economic conditions stabilized, instrumentation gradually shifted towards inflation-targeting principles. This process culminated in 2001, when the Bank of Mexico officially adopted a fully-fledged inflation-targeting framework. Thus, if monetary policy has effectively contributed to lower inflation, then this suggests that inflation-targeting in emerging market economies is a useful mechanism for imposing discipline on monetary policy and ensuring that it performs the role of nominal anchor of the economy. The rest of the paper is organized in four sections. Section 2 reviews the literature on monetary policy rules and shows that under a flexible exchange-rate regime this framework is useful in analyzing the role of monetary policy as the nominal anchor of the economy. Section 3 specifies a baseline monetary policy rule and describes the estimation procedure. Results from the baseline exercise are presented in this section and are used to motivate the exercises in the following section. Section 4 specifies and estimates alternative monetary policy rules and tests the role of several variables in the process through which interest rates have been determined in Mexico. Section 5 concludes.

2. Analyzing monetary policy through monetary policy rules A key question about the conduct of monetary policy is whether it should follow a particular rule or be discretionary. The current consensus is that rules are needed to prevent time-inconsistent discretionary policies from causing the problem of “inflation bias”.1 1

See Kydland and Prescott (1977), Barro and Gordon (1983), and Rogoff (1985).

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

359

Following Taylor (1993, 1999), a monetary policy rule is best described as a systematic approach to analyze the instrumentation of monetary policy. The first step in specifying a monetary policy rule is to choose a variable to represent the instrument of monetary policy. Then, it is necessary to choose the set of variables to which monetary policy will react, that is, variables which will induce the central bank to modify the monetary policy instrument. In the rule set forth by Taylor (1993), the nominal interest rate is the instrument of monetary policy and the deviation of inflation from its target and the output gap are the variables to which it would react: it = α + β(πt − π∗ ) + γ(yt − yt∗ ),

(1)

where it stands for the nominal interest rate, πt is the inflation rate,2 π∗ is the inflation target, yt is output, and yt∗ represents potential output. The parameter α represents the long-run equilibrium nominal interest rate, while parameters β and γ measure the magnitude of the hypothetical response of the monetary policy instrument to deviations of inflation from its target and to the output gap, respectively. An issue subject to considerable discussion is the role of the output gap in monetary policy rules. A central bank which has price stability as its main objective, could choose to adjust its policy instrument only in response to deviations of inflation from its target (no output gap in the policy rule). This approach is known as extreme inflation targeting. To understand the difference between an inflation-targeting rule like Eq. (1) and extreme inflation targeting, consider the following two cases. First, when inflationary pressures are due to an excess of aggregate demand, interest rates should rise according to Eq. (1), since both inflation and the output gap increase. Under an extreme inflation-targeting rule, the recommendation would be the same, namely, to raise interest rates in response to the increase in inflation. Therefore, when inflationary pressures come from the demand side there is no qualitative difference between a monetary policy rule like Eq. (1) and extreme inflation targeting. Consider now the case where inflationary pressures come from an adverse cost-push shock. Under extreme inflation targeting, the transitory rise in the price of inputs and the resulting increase in inflation induces monetary policy to raise interest rates in order to prevent prices of final goods from rising. The result would be a contraction of economic activity. However, when the output gap is included in the monetary policy rule, as in Eq. (1), the higher interest rate required by the rise in inflation is compensated with a lower interest rate needed by the contraction in the output gap. Thus, if interest rates rise at all, they rise by a smaller magnitude than under extreme inflation targeting and the adverse effect on economic activity is smaller. Therefore, a monetary policy rule which includes both inflation and the output gap is likely to offer a better approximation to the process through which a central bank, following inflation-targeting principles, sets interest rates. That is, a central bank that lets adverse transitory cost-push shocks have a first-order effect (once-and-for-all) on prices, but that immediately fights inflationary pressures originating on the demand side.3 2 π represents the inflation rate between time “t − n” and “t”, where typically “n” stands for 1 year. Thus, t although πt is known at time “t”, it captures the inflation rate observed from the past to present. 3 See Clarida et al. (1999).

360

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

The theoretical foundations of the rule proposed by Taylor have been formalized by several authors. Svensson (1996), Clarida, Gal´ı, and Gertler (1999) and Woodford (2001) have shown that approximations to this rule could be obtained from an optimization process of a central bank that minimizes a quadratic loss function in inflation deviations from its target and the output gap, subject to a standard New Keynesian macroeconomic framework. In particular, they postulate a forward-looking version of the rule proposed by Taylor: ∗ ]), it = α + β(Et [πt+n − π∗ ]) + γ(Et [yt+k − yt+k

(2)

where πt +n represents inflation between period t and period t + n, yt +k is output at period t + k, and Et is the expectations operator with information available at time t. Interest rates are assumed to respond to expected inflation deviation from its target and to the expected output gap, rather than with respect to observed changes in the past and present values of these variables. For a monetary policy rule like Eq. (2) to be consistent with a central bank that minimizes a quadratic loss function on inflation deviations from its target and the output gap, parameters β and γ have to be greater than 1 and 0, respectively.4 To see this, consider first parameter β which describes the magnitude of the response of interest rates, when inflation expectations are above or below the target. Given that α represents the long-run equilibrium nominal interest rate, the monetary policy rule can be expressed as follows: ∗ ]), rt = r¯ + (β − 1)(Et [πt+n − π∗ ]) + γ(Et [yt+k − yt+k

(3)

where rt is the real interest rate and r¯ is the long-run level of the real interest rate. From Eq. (3) it is clear that the critical value of parameter β is 1. Consider, for example, the case where inflation expectations rise above the inflation target (Et [pt+n ] > π∗ ). When β > 1, the rule suggests that nominal interest rates (it ) rise sufficiently to let the real interest rate (rt ) increase as well. The effect of this adjustment contracts aggregate demand and induces inflation expectations to converge to the target. This case would represent a restrictive monetary policy adjustment. When 0 < β < 1, the increment in nominal interest rates (it ) is not enough to induce the real interest rate (rt ) to rise. In this case, the real interest rate (rt ) not only does not rise, but it decreases since the nominal interest rate (it ) increment is smaller than the rise in expected inflation Et [πt+n ]. The reduction of the real interest rate stimulates aggregate demand and fuels inflation expectations. If a monetary policy rule like Eq. (2) with β > 1 offers a good approximation to the process through which interest rates are determined, then it is safe to say that monetary policy works as an automatic stabilizer of inflation around its target, that is, it serves as nominal anchor of the economy. Whenever inflation expectations deviate from the target, a rule like this would require the central bank to act so as to ensure the convergence of expected inflation, and eventually of actual inflation, to the inflation target. In this case, the process through which interest rates are determined is consistent with a central bank that follows inflation-targeting principles. With respect to parameter γ, Eqs. (2) and (3) show that the critical value is 0. Consider ∗ ] > 0). the case where output is expected to be above its potential level (Et [yt+k − yt+k When γ > 0, the response implied by the rule is to raise nominal (it ) and real (rt ) interest 4

See Woodford (2001).

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

361

rates in order to contract aggregate demand, induce output to converge to its potential level and avoid pressures on future inflation. When γ < 0, the rule implies that the reduction in nominal (it ) and real (rt ) interest rates fuels aggregate demand and pushes output farther above its potential level, which eventually will cause inflation to rise. Once again, when a rule like Eq. (2) with γ > 0 offers a good approximation to the process through which interest rates are determined, monetary policy is said to be an automatic stabilizer of output around its potential level. This type of policy is consistent with inflation targeting, since it eliminates sources of persistent pressure on inflation. The stabilizing properties of monetary policy, when it can be described by a monetary policy rule with parameters β > 1 and γ > 0, suggest that monetary policy not only works as a nominal anchor, but also promotes a stable macroeconomic background in which output growth is primarily determined by technology and other supply-side elements (potential output). So far, the discussion has not considered the special problems faced by central banks in small open economies, including volatility in external financial markets and exchange-rate shocks. While several authors have extended the analysis of monetary policy rules to small open economies, there is no consensus on how monetary policy rules should be specified. Ball (1999) and Svensson (2000) argue that monetary policy rules should encompass other variables, such as foreign interest rates or the exchange rate, for example, which may under some circumstances reflect uncertainties about future inflation or the expected output gap. A representation of this type of rule is given by ∗ it = α + β(Et [πt+n − π∗ ]) + γ(Et [yt+k − yt+k ]) + ϕ(Et [zt+m ]),

(4)

where zt +m could represent the exchange rate, money, foreign interest rates, country-risk perception or any other variable which may have an influence in the process through which interest rates are determined. On the other hand, Clarida, Gal´ı, and Gertler (2001) present a small, open-economy model in which the monetary policy rule is similar to the one considered for the closed economy (Eq. (2)), except that the effect of the exchange rate on output and inflation is captured in the parameters of the policy rule. The response of interest rates is thus qualitatively the same in closed and open economies. Interest rates are adjusted in response to expected inflation and the output gap. The difference is in the quantitative response (magnitude of parameters β and γ) of interest rates to expected inflation and the output gap. Openness affects the parameters of the model, but not the general form of the policy rule. The intuition behind this argument is that if movements in the exchange rate or other variables are to have an effect on inflation and/or output, then these effects should be captured by changes in the expected deviation of inflation from its target and in the expected output gap. However, if movements in these other variables are not expected to have an effect on either inflation or output, then there is no need to adjust interest rates. The context in which monetary policy rules have been discussed, provides an analytical framework in which to analyze two interesting issues related to monetary policy in Mexico. First, since the adoption of the flexible exchange-rate regime in 1994, monetary policy is, in theory, the nominal anchor of the economy. A formal test of whether the Bank of Mexico has succeeded in performing this role involves the identification of a monetary policy rule, that offers a good approximation to the process through which interest rates are determined in

362

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

Mexico, and then to test whether parameter β is greater than 1. A second issue is whether, in addition to being the nominal anchor of the economy, monetary policy has followed inflation-targeting principles rather than extreme inflation targeting? For this to be the case, in addition to β being greater than 1, it must also be the case that γ is greater than 0.

3. A monetary policy rule for Mexico: baseline case This section follows the literature5 in estimating a baseline monetary policy rule to approximate the process through which interest rates are set in Mexico. The Bank of Mexico does not determine the level of interest rates directly by announcing a target and then pursuing that level through open-market operations.6 It is rather through the conditions under which commercial banks hold current accounts in the Central Bank (overdraft facilities, remuneration of deposits and penalty rates) that the Bank of Mexico signals changes in the stance of monetary policy and lets interest rates be adjusted by the market. If the resulting interest rate levels are consistent with the inflation target, then the Bank has no need to signal a change in the stance of monetary policy. However, if interest rates are not consistent with the inflation target, then the Bank modifies the conditions until interest rates reach a level which satisfies the Central Bank. Thus, although interest rates are not directly determined by the Bank of Mexico, their behavior is steered by the stance of monetary policy. Hence, we seek to identify a monetary policy rule that approximates the process through which interest rates are determined and thereby to characterize monetary policy in Mexico. 3.1. Specification Continuing the discussion of the previous section, we assume that a monetary policy rule can be described by ∗ ∗ ∗ ) + β(Et [πt+n − πt+n ]) + γ(Et [yt+k − yt+k ]), i∗t = (κ + απt+n

(5)

∗ repwhere i∗t represents the interest-rate target (monetary policy instrument); κ + απt+n resents the long-run nominal interest rate; and all the other variables and parameters are defined as before. In this case, it is important to note that the long-run nominal interest ∗ ) and the inflation target (π ∗ ) are not assumed to be constant over time. rate (κ + απt+n t+n This feature takes account of the fact that, in recent years, the Mexican economy has been through a disinflation process in which the inflation targets set by the Bank of Mexico have been decreasing. Another practical issue is the speed with which actual interest rates (it ) converge to targeted interest rates (i∗t ). Following Clarida, Gal´ı, and Gertler (1998, 2000), it is assumed that actual interest rates are adjusted slowly. Thus, at any given time, actual interest rates (it ) are measured as a weighted average of the targeted interest rate (i∗t ), determined by Eq. (5) 5 6

See Clarida et al. (1998). See Banco de M´exico (2002) Schwartz and Torres (2002), and Mart´ınez, Sanchez and Werner (2001).

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

363

and the observed interest rate in the previous period (it −1 ), plus an exogenous interest rate shock (vt ) with zero mean:7 it = (1 − ρ)i∗t + ρit−1 + vt ,

(6)

where parameter ρ takes values between 0 and 1 and measures the degree of interest-rate smoothing. Finally, the combination of Eqs. (5) and (6) provides the baseline monetary policy rule to be estimated: ∗ ∗ it = (1 − ρ)(κ + απt+n ) + (1 − ρ)β(Et [πt+n − πt+n ]) ∗ ]) + ρit−1 + vt . + (1 − ρ)γ(Et [yt+k − yt+k

(7)

3.2. Estimation technique The estimation of Eq. (7) is not straight-forward, since it includes as right-hand side variables the expected inflation deviation from its target and the expected output gap. First, information regarding the central bank’s inflation forecasts is not available in Mexico. Nevertheless, the inflation-deviation term could be constructed with the information reported from market surveys about expected future inflation.8 However, relying exclusively on the market’s expected inflation has its shortcomings. If market inflation expectations rise by a smaller margin than the central bank’s inflation forecasts, then the estimated β would be smaller than the one obtained if the central bank’s forecasts were used to model expected inflation, and would be interpreted as a less aggressive change in interest rates. Second, information regarding the central bank’s expected output gap is also not available. To get around these two problems Clarida, Gal´ı and Gertler (1998, 1999, 2000) propose the generalized method of moments (GMM) to estimate Eq. (7).9 The GMM estimation technique is best explained by noting that Eq. (7) could be expressed as follows: ∗ ∗ it = (1 − ρ)(κ + απt+n ) + (1−ρ)β(πt+n − πt+n ) + (1−ρ)γ(xt+k ) + ρit−1 + εt ,

(8) ∗ ) is the output gap and the error term ε is defined as where xt+k = (yt+k − yt+k t ∗ ∗ εt = vt − (1 − ρ)β{(πt+n − πt+n ) − Et [πt+n − πt+n ]}

− (1 − ρ)γ(xt+k − Et [xt+k ]).

(9)

Eq. (8) implies that ex-post observed values of inflation deviation from its target (πt+n − ∗ ) and of the ex-post output gap (x πt+n t +k ) are used to approximate the expected inflation 7 This type of shock could capture the effect of a random (not an endogenous response of monetary policy) change in the stance of monetary policy or a temporary adjustment in interest rates not explained by output or inflation. 8 See Infosel (2002): “Encuesta Semanal sobre las Expectativas de los Especialistas en Econom´ıa del Sector Privado”. 9 Notation follows Clarida et al. (1998).

364

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

∗ ]) and the expected output gap (E [x deviation from its target (Et [πt+n − πt+n t t +k ]), respectively. Notice that the error term εt is a linear combination of the exogenous monetary policy shock (vt ) and the forecast errors. Then, using ex-post data will deliver unbiased estimators of parameters κ, α, β, γ and ρ if the forecast errors have a zero mean. Thus, in order to find a set of parameters which will guarantee average zero forecast errors, GMM uses the information contained in a set of variables (Ut ) known at time t when the interest rate it is determined. This set of instrument variables, Ut , typically includes lagged values of inflation, output, interest rates and, in general, any other variable which is useful in forecasting inflation and output. This strategy imposes a set of orthogonal restrictions, which are used by GMM to estimate parameters κ, α, β, γ and ρ that are given by E[εt |Ut ] = 0.10 Combining this condition with Eq. (8), the explicit set of restrictions exploited by GMM is ∗ ∗ E[it − (1 − ρ)(κ + απt+n ) − (1 − ρ)β(πt+n − πt+n )

− (1 − ρ)γ(xt+k ) − ρit−1 |Ut ] = 0.

(10)

3.3. Sample period and definition of variables As mentioned before, since Mexico adopted a flexible exchange-rate arrangement late in 1994, the Bank of Mexico has had the responsibility of establishing monetary policy as the nominal anchor of the economy. Over this period, inflation rose during the first part of 1995 and then started to decrease gradually, suggesting that monetary policy has been successful in bringing inflation down. However, it could also be that inflation has come down due to other favorable factors and that monetary policy has played only an accommodative role. Therefore, it is necessary to choose a sample period during which the monetary authorities were “tested”, that is, a period during which unfavorable domestic or international influences required assertive monetary policy actions to prevent a permanent rise in inflation. Figs. 1 and 2 show data on inflation, interest rate, country-risk perception and nominal exchange rate in Mexico from 1996 to 2001. In Fig. 1, the solid line represents annual CPI inflation, while the dotted line gives the monthly average of money-market daily overnight interest rates. Fig. 2 shows country-risk perception, represented by the “EMBI” spread for Mexico (solid line)11 and the nominal exchange rate (dotted line). During 1996 and most of 1997, inflation and country-risk perception followed a downward trend, which was accompanied by a decline in interest rates. Up to September of 1997, a favorable international environment allowed the Bank of Mexico to benefit from the declining inflation rate without incurring the cost of rising interest rates. However, uncertainty in international financial markets increased as a result of the Asian financial problems and in October of 1997 (vertical line) country-risk perception for most emerging markets, Mexico included, started to increase sharply. This event clearly represented an adverse shock to Mexico and other emerging market economies. However, despite 10 When the number of instruments (variables included in U ) is larger than the number of parameters to be t estimated, the model is over-identified. In this case, Hansen’s J-test is commonly used to test if the over-identifying restrictions hold. 11 The Emerging Market Bond Index spread is obtained from J.P. Morgan.

Fig. 2. Country risk and exchange rate.

10 / 01

07 / 01

04 / 01

01 / 01

10 / 00

07 / 00

04 / 00

01 / 00

10 / 99

07 / 99

6.5

04 / 99

01 / 99

10 / 98

07 / 98

04 / 98

01 / 98

10 / 97

07 / 97

04 / 97

01 / 97

10 / 96

07 / 96

04 / 96

01 / 96

Nominal exchange rate (pesos / dollar) 9.0 800

8.5

8.0 600

7.5

400

7.0

Nominal exchange rate

200

Country risk

6.0

0

Contry risk (basis points)

10.5 11 / 01

09 / 01

07 / 01

05 / 01

03 / 01

01 / 01

11 / 00

09 / 00

07 / 00

05 / 00

03 / 00

01 / 00

11 / 99

09 / 99

07 / 99

05 / 99

03 / 99

01 / 99

11 / 98

09 / 98

07 / 98

05 / 98

03 / 98

01 / 98

11 / 97

09 / 97

07 / 97

05 / 97

03 / 97

01 / 97

11 / 96

09 / 96

07 / 96

05 / 96

03 / 96

01 / 96

%

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379 365

60

Inflation

50

Interest rate

40

30

20

10

0

Fig. 1. Inflation and interest rate.

1200

10.0

9.5 1000

366

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379 30

Inflation 25

Inflation target

%

20

15

10

5

10 / 01

07 / 01

04 / 01

01 / 01

10 / 00

07 / 00

04 / 00

01 / 00

10 / 99

07 / 99

04 / 99

01 / 99

10 / 98

07 / 98

04 / 98

01 / 98

10 / 97

07 / 97

04 / 97

01 / 97

0

Fig. 3. Inflation and inflation target.

a temporary rise in the inflation rate at the end of 1998 and an exchange-rate depreciation, the inflation-reduction process resumed and one-digit inflation rates have been observed in Mexico since late 2000. We test the hypothesis that monetary policy has played the role of nominal anchor in the Mexican economy by examining the behavior of interest rates after October 1997. To test this hypothesis, the baseline monetary policy rule (Eq. (7)) is estimated to determine whether β is statistically greater than 1. The sample period starts in October of 1997; n is set equal to 12 and k is set equal to 0. This combination implies that interest rates depend on 1-year-ahead expected inflation and on the contemporaneous expected output gap. The interest rate (it ) in Eq. (7) is represented by the monthly average of money-market daily overnight rates, as in Fig. 1. As indicated in Eq. (8), the expected inflation deviation ∗ from its target (Et [πt+12 − πt+12 ]) in Eq. (7) is approximated by the ex-post observed ∗ inflation deviation from its target (πt+12 − πt+12 ). Observed inflation is represented by 12-month CPI inflation and the annual inflation target at each month is based on the annual inflation targets announced by the Bank of Mexico. Fig. 3 shows the ex-post observed inflation rate (solid line) and the inflation target (dotted line) for the sample period. As given in Eq. (8), the expected output gap (Et [yt − yt∗ ]) is approximated by the ex-post observed output gap (yt − yt∗ ), based on the Index of Economic Activity (IGAE).12 Then, following 12

Two other variables were considered: an Index of Industrial Production (IVPI) and an estimated measure of monthly GDP. Results are not affected by the output definition.

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

367

4.90

Output 4.85

Potential output

log

4.80

4.75

4.70

10 / 01

07 / 01

04 / 01

01 / 01

10 / 00

07 / 00

04 / 00

01 / 00

10 / 99

07 / 99

04 / 99

01 / 99

10 / 98

07 / 98

04 / 98

01 / 98

10 / 97

07 / 97

04 / 97

01 / 97

4.65

Fig. 4. Output and potential output.

standard procedures in the literature, potential output is estimated using a quadratic trend.13 Fig. 4 shows observed and potential output. 3.4. Results for the baseline case The baseline monetary policy rule described by Eq. (7) is estimated using GMM for three different sets of instruments, Ut . As noted, the instruments should typically include information available when interest rates were set and should be useful in estimating the expected inflation deviation from its target and the expected output gap. The first set of ∗ ), the output instruments includes lags 1–6, 9, and 12 of the inflation deviation (πt−j − πt−j gap (xt −j ) and the interest rate (it −j ).14 Results are presented in the first line in Table 1. The fact that none of the parameters, except for ρ, are statistically different from 0 suggests that the information included in the set of instruments (past inflation deviations, past output gap and past interest rates) may not be good enough to adequately forecast inflation and the output gap. Therefore, the set of instruments is enriched with two variables which are likely to have a relationship with future inflation and with the output gap. The second set (case 2) of instruments includes lagged information on nominal wages from the 13 Potential output was also estimated using linear and Hodrick–Prescott trends. Results do not vary significantly from those reported under a quadratic trend. 14 Results do not vary significantly from those reported when the set of instruments includes lags 1–6 or 1, 3, 6, 9, and 12.

368

Instruments Ut Case 1: Case 2: Case 3: a

∗ ,x Ut : {1, πt−j − πt−j t−j , it−j }, for j = 1–6, 9, 12 ∗ ,x Ut : {1, πt−j − πt−j t−j , it−j , wt−j }, for j = 1–6, 9, 12 ∗ ,x Ut : {1, πt−j − πt−j t−j , it−j , st−j }, for j = 1–6, 9, 12

κ −20.4 (29.9) 0.96 (11.7) 9.27 (8.6)

α 3.41 (2.80) 1.53 (1.07) 0.80 (0.78)

β 2.31 (1.42) 2.18∗∗ (0.68) 1.99∗∗ (0.56)

γ

ρ

J-test

6.46 (3.45) 3.57∗∗ (1.60) 2.12∗ (1.15)

0.93∗∗

4.92 (0.99) 5.13 (0.99) 5.12 (0.99)

(0.02) 0.92∗∗ (0.01) 0.92∗∗ (0.01)

Standard deviations in parentheses, except in the J-test column, where they represent the “P” value to reject the hypothesis that over-identifying restrictions hold. Statistically significant at 90% confidence level. ∗∗ Statistically significant at 95% confidence level. ∗

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

Table 1 Monetary policy rule: baseline estimationa

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

369

manufacturing sector (wt ), while the third set (case 3) includes lagged information on the nominal exchange rate (st ).15 Results are reported in lines 2 and 3 in Table 1. In cases 2 and 3, parameters β and γ are statistically significant at the 95% confidence level, except for γ in case 3, which is significant only at the 90% confidence level.16 In both cases, parameters κ and α, which are related to the long-run nominal interest rate, are not statistically significant.17 It is not possible to reject the hypothesis that all the over-identifying restrictions are satisfied. Thus, the set of instruments used seems to be appropriate in both cases. The differences in results between case 1 and cases 2 and 3 suggest that the information contained in nominal wages and in the nominal exchange rate is useful in forecasting inflation and output and consequently improves the adjustment of the policy rule. The key result in this exercise is the magnitude of parameter β. As mentioned before, when β > 1, the policy rule implies that if expected inflation is above the inflation target, then the nominal interest rate rises sufficiently to induce an increment in the real interest rate, and thus monetary policy becomes the nominal anchor of the economy. In cases 2 and 3, the estimate of parameter β is positive and greater than 1. This result is confirmed with a formal test of whether the coefficient estimate is statistically greater than 1. The “t” statistics for cases 2 and 3 are 1.73 and 1.76, respectively. Thus, the null hypothesis that β ≤ 1 is rejected at the 95% confidence level in both cases (critical value for one side test is 1.64). This result suggests that monetary policy in Mexico, through its effect on interest rates, has effectively played the role of nominal anchor for the economy. Another noteworthy result is that the estimates of parameter γ are positive and significant. This suggests that movements in interest rates, while allowing output to fluctuate around its potential level, have prevented the presence of permanent inflationary pressures. This reinforces the hypothesis that by using interest rates to fight demand-side sources of inflation, monetary policy in Mexico has become the nominal anchor of the economy. Furthermore, the fact that β > 1 and γ > 0 suggests that monetary policy has been consistent with inflation-targeting principles. Another way to evaluate the baseline monetary policy rule, is to compare the estimated “policy interest rate” with actual interest rates, that is, to compare observed interest rates with those that would have been observed if during the sample period interest rates had been determined by the estimated policy rule. Fig. 5 shows the actual interest rate (solid line) and the simulated or dynamic “policy interest rate” (dotted line).18 Overall, the “policy interest rate” does a good job in following the direction of actual interest rates. However, there are important deviations, which suggest that the baseline policy rule is not the best representation of how interest rates are determined. For example, as a result of the increasing gap between inflation expectations and the inflation target late in 1997 and early 1998 (Fig. 3), the policy 15 In order to have stationary variables, both wages and nominal exchange rates were defined as the first log difference. 16 Rodr´ıguez (2001) and Mart´ınez et al. (2001) estimate similar specifications of monetary policy rules for Mexico after 1996 and also find that the output gap has a lower level of significance than the term which captures the deviation of inflation from its target. 17 Even though these parameters are not statistically significant, their inclusion in the specification prevents parameters β and γ from being biased, because of the decreasing trend of the interest rate over the sample period. 18 The “policy interest rates” from cases 2 and 3 are almost identical. Case 2 is presented in Fig. 5.

370

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379 40

Actual interest rate

35

30

Dynamic "policy interest rate"

%

25

20

15

10

5

12 / 00

10 / 00

08 / 00

06 / 00

04 / 00

02 / 00

12 / 99

10 / 99

08 / 99

06 / 99

04 / 99

02 / 99

12 / 98

10 / 98

08 / 98

06 / 98

04 / 98

02 / 98

12 / 97

10 / 97

0

Fig. 5. Interest rate and dynamic “policy interest rate”.

rule suggested a higher interest rate than the one observed. On the other hand, late in 1998 and early 1999 the policy rule suggested an interest rate below actual interest rates, reflecting inflation expectations below the target and a negative output gap (Figs. 3 and 4). Results from the baseline case suggest that monetary policy in Mexico has been consistent with inflation-targeting principles and thus has performed the role of nominal anchor of the economy. However, so far these conclusions should be taken with reserve, since the base line rule does not seem to be a good representation of how interest rates are determined in Mexico.

4. Augmented monetary policy rules for Mexico It is possible that in a small open economy like Mexico’s, other variables need to be considered in order to have a better approximation to the process through which interest rates are determined. In what follows, an augmented monetary policy rule is used to test the role of macroeconomic variables other than inflation and output. In this section, it is assumed that the augmented monetary policy rule is described by ∗ ∗ ∗ i∗t = (κ + απt+n ) + β(Et [πt+n − πt+n ]) + γ(Et [yt+k − yt+k ]) + ϕ(Et [zt+m ]),

(11) where zt +m represents any variable, other than inflation and output, that may influence the process through which interest rates are determined. Then, combining this policy rule with

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

371

the interest-rate-smoothing equation (Eq. (6)), implies that the policy rule to be estimated is ∗ ∗ ) + (1 − ρ)β(Et [πt+n − πt+n ]) it = (1−ρ)(κ+απt+n

∗ ]) + (1 − ρ)ϕ(Et [zt+m ]) + ρit−1 + vt , + (1 − ρ)γ(Et [yt+k − yt+k

(12)

where variable zt +m is defined explicitly in the following subsections. 4.1. The forward-looking component of monetary policy In recent years, instrumentation of monetary policy in Mexico has evolved towards inflation-targeting principles. The relevant literature suggests that the monetary authorities should be forward-looking and that decisions should be related to the expected future performance of the economy.19 Thus, as monetary policy becomes more forward-looking, expected inflation becomes more important and lagged inflation less important in the determination of interest rates. Before testing the role of additional macroeconomic variables in the monetary policy rule, in this exercise we test whether interest rates are determined more in a forward-looking than backward-looking manner. The augmented monetary policy rule (Eq. (12)) is estimated and variable zt +m is defined as the observed lagged inflation ∗ ).20 If interest rates are determined in a relatively backward-looking deviation (πt−1 − πt−1 manner, the nominal interest rate would be expected to rise enough to raise the real interest rate when lagged inflation is above its target (ϕ > 1), rather than when inflation is expected to be above its target (β > 1). The results, presented in Table 2,21 suggest that, in line with inflation-targeting principles, the process through which interest rates are determined in Mexico is “more” forward-looking than backward-looking. Parameter β is statistically greater than 1 while parameter ϕ is not statistically different from 0. This implies that an increase in the nominal interest rate produces an increase in the real interest rate in response to expected inflation, but not in response to lagged inflation. 4.2. Interest rates, money and the exchange rate In this exercise we test whether including money or the exchange rate in the augmented policy rule helps to better approximate the process by which interest rates are determined. In the first exercise, variable zt +m in the augmented monetary policy rule (Eq. (12)) is defined as the monthly variation (first log difference) of the monetary base (mbt ). Results reported in the first line in Table 3 show that parameter ϕ is positive and statistically different from 0.22 However, β is no longer greater than 1. Similarly, in the second exercise (second row in Table 3), where variable zt +m is defined as the monthly variation (log first difference) of the nominal exchange rate expressed in pesos per dollar (ert ), 19

See Bernanke, Laubach, Mishkin, and Posen (1999). Results do not vary significantly when the lagged inflation deviation is considered in periods t − 6 or t − 12. 21 As in the baseline exercise, results are robust to different combinations of lags for the set of instruments included in Ut . 22 Results for the two exercises reported in Table 3 are robust to different combinations of lags for the set of instruments included in Ut . 20

372

Variable z; instruments Ut

κ

α

β

γ

ϕ

ρ

J-test

∗ ; U : {1, π ∗ zt+m = πt−1 − πt−1 t t−j − πt−j , xt−j , it−j }, for j = 1–6, 9, 12

−16.8 (14.4)

3.51∗∗ (1.42)

1.57∗∗ (0.27)

2.59∗ (1.52)

−1.80 (1.14)

0.93∗∗ (0.01)

5.12 (0.99)

a

Standard deviations in parentheses, except in the J-test column, where they represent the “P” value to reject the hypothesis that over-identifying restrictions hold. Statistically significant at 90% confidence level. ∗∗ Statistically significant at 95% confidence level. ∗

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

Table 2 Augmented monetary policy rule: the forward-looking componenta

Variable z; instruments Ut

κ

α

β

γ

ϕ

ρ

J-test

zt+m = mbt ; ∗ ,x Ut : {1, πt−j − πt−j t−j , it−j , wt−j , zt−l }, for j = 1–6, 9, 12 and for l = 1–3 zt+m = ert ; ∗ ,x Ut : {1, πt−j − πt−j t−j , it−j , wt−j , zt−l }, for j = 1–6, 9, 12 and for l = 1–3

−28.3∗ (15.79)

4.20∗∗ (1.42)

0.66∗∗ (0.40)

6.18∗∗ (2.03)

0.74∗∗ (0.20)

0.92∗∗ (0.01)

4.90 (0.99)

1.27∗ (0.75)

0.67∗∗ (0.27)

1.11 (0.82)

1.62∗∗ (0.45)

0.87∗∗ (0.02)

5.05 (0.99)

a

4.34 (8.14)

Standard deviations in parentheses, except in the J-test column, where they represent the “P” value to reject the hypothesis that over-identifying restrictions hold. Statistically significant at 90% confidence level. ∗∗ Statistically significant at 95% confidence level. ∗

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

Table 3 Interest rates, money and the exchange ratea

373

374

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379 40

Actual interest rate

35

Dynamic "policy interest rate" with money

30

Dynamic "policy interest rate" with exchange rate

%

25

20

15

10

5

12 / 00

10 / 00

08 / 00

06 / 00

04 / 00

02 / 00

12 / 99

10 / 99

08 / 99

06 / 99

04 / 99

02 / 99

12 / 98

10 / 98

08 / 98

06 / 98

04 / 98

02 / 98

12 / 97

10 / 97

0

Fig. 6. Interest rate and dynamic “policy interest rates”.

β is not greater than 1 and γ is statistically not different from 0. It seems that part of the information contained in either of these two variables is also incorporated in expected inflation and the expected output gap and thus the magnitude and significance level of β and γ decreases. These results suggest that a monetary policy rule which simultaneously includes expected inflation, the expected output gap and money or the exchange rate may not be the best approximation to the process by which interest rates are determined in Mexico. This is evident in Fig. 6 where the actual interest rate and the dynamic “policy interest rate” from each of the two specifications are compared. In both cases the rules generally follow the direction of actual interest rates. However, as in the baseline monetary policy rule (Fig. 5) in late 1997 and early 1998, deviations from the actual interest rate are considerable. 4.3. Country-risk perception, interest rate differential and the monetary policy rule In this exercise, another two variables are considered in the augmented monetary policy rule. The first is country-risk perception. As seen in Fig. 2, the financial problems experienced in Asia, Russia and Latin America in recent years had negative effects on the country-risk perception of Mexico and other emerging markets. Therefore, by considering the effect of country risk in the monetary policy rule, the aim is to incorporate influences stemming from international financial markets. The second variable to be considered is the differential between long-run and short-run domestic interest rates. In this case, the purpose

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

375

is to incorporate the uncertainty created by domestic events, such as the presidential election of 2000, which generated temporary uncertainty and which was not reflected in changes in country-risk perception. For country-risk perception, variable zt +m in the augmented policy rule (Eq. (12)) is defined as the monthly average of the “EMBI” spread for Mexico observed in the previous period (crt −1 ).23 For the interest-rate differential, variable zt +m is defined as the lagged difference between the monthly average of 1-year treasury bills (Cetes) and the monthly average of the overnight money-market interest rate (dit −1 ).24 Results for the two cases are presented in the first two lines in Table 4, respectively.25 As in the baseline case, parameter β is statistically larger than 1 and parameter γ is statistically larger than 0. In both cases, also, parameter ϕ is statistically different from 0 and positive. This result shows that both variables have a positive effect on the interest rate, thereby capturing the effect on interest rates of both external and domestic uncertainty, respectively. A third specification of the augmented monetary policy rule combines country-risk perception and the interest-rate differential as variables z1t+m and z2t+m , respectively. Results in the third line in Table 4 confirm previous findings, parameters β and γ being statistically greater than 1 and 0, respectively. Furthermore, when taken simultaneously the two variables are important in the determination of interest rates since parameters ϕ1 and ϕ2 are positive and statistically different from 0. Thus, whenever country-risk perception rises, domestic interest rates rise as well; and whenever the difference between long- and short-run domestic interest rates increases, the monetary policy rule suggests an increase in the short-run domestic interest rate. The performance of the augmented policy rule, which includes both country-risk perception and the interest-rate differential, is presented in Fig. 7. In this case, the dynamic simulation seems to follow closely the performance of actual interest rates. It is remarkable that the “policy interest rate” is able to follow the sudden climb in actual interest rates at the end of 1998 and then to follow the reduction early in 1999. Despite a temporary deviation from the actual interest rate in late 1999, the “policy interest rate” resumes its downward path, just as actual interest rates do. The evidence presented shows that the augmented policy rule, which includes information on the expected inflation deviation from its target, the expected output gap, the perception of country risk and the difference between long- and short-run interest rates, provides a good approximation to the process through which interest rates are determined in the Mexican economy. The evidence also suggests that the monetary policy instrumented by the Bank of Mexico in recent years has been in line with inflation-targeting principles and consequently has performed the role of nominal anchor of the economy. 23

The “EMBI” is expressed in percentage points rather than basis points. The 1-year treasury bill (Cetes 364) was not issued from September to December 1998. For those dates the 3-month treasury bill (Cetes 91) converted to a 364-day curve is considered as the long-run interest rate. The differential is expressed in percentage points. 25 Results for the three exercises reported in Table 4 are robust to different combinations of lags for the set of instruments included in Ut . 24

376

Variable z; instruments Ut zt+m = crt−1 ; ∗ Ut : {1, πt−j − πt−j , xt−j , it−j , wt−j , zt−l }, for j = 1–6, 9, 12 and for l = 1–3 zt+m = dit−1 ; ∗ Ut: {1, πt−j − πt−j , xt−j , it−j , wt−j , zt−l }, for j = 1–6, 9, 12 for l = 1–3 z1t+m = crt−1 , z2t+m = dit−1 ; Ut : {1,πt−j − π∗t−j , xt−j , it−j , wt−j , z1t−l }, for j = 1–6, 9, 12 and for l = 1–3 a

κ 12.6

α ∗∗

(2.77)

−1.61

β ∗∗

(0.29)

1.91

γ ∗∗

(0.12)

1.08

ϕ ∗∗

(0.24)

33.4 (21.7)

−4.91 (3.31)

6.90∗∗ (3.03)

5.74∗∗ (1.98)

−2.18 (1.21)

−0.05 (0.17)

1.35∗∗ (0.12)

1.68∗∗ (0.15)

0.72

ϕ1 ∗∗

ϕ2

(0.14)

ρ 0.70

11.7∗∗ (4.87)

3.21∗∗ (1.56)

0.27∗∗ (0.10)

(0.01)

5.19 (0.99)

0.96∗∗ (0.01)

5.22 (0.99)

0.67∗∗ (0.01)

4.86 (0.99)

Standard deviations in parentheses, except in the J-test column, where they represent the “P” value to reject the hypothesis that over-identifying restrictions hold. Statistically significant at 90% confidence level. ∗∗ Statistically significant at 95% confidence level. ∗

J-test ∗∗

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

Table 4 Augmented monetary policy rule: country-risk and interest-rate differentiala

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

377

40

35

Actual interest rate

30

Dynamic "policy interest rate"

%

25

20

15

10

5

12 / 00

10 / 00

08 / 00

06 / 00

04 / 00

02 / 00

12 / 99

10 / 99

08 / 99

06 / 99

04 / 99

02 / 99

12 / 98

10 / 98

08 / 98

06 / 98

04 / 98

02 / 98

12 / 97

10 / 97

0

Fig. 7. Interest rate and dynamic “policy interest rate”.

5. Conclusions This paper analyzes the process through which interest rates are determined in the Mexican economy. It focuses on two fundamental issues of monetary policy. The first seeks to ascertain whether in the transition towards an inflation-targeting framework, monetary policy has become the nominal anchor of the Mexican economy? The second asks whether the instrumentation of monetary policy has been in line with inflation-targeting principles? While analyzing these two issues, we identify a set of variables that, when combined within a monetary policy rule, provide a good approximation to the process through which interest rates are determined. The evidence presented shows that in recent years monetary policy in Mexico, through its effect on interest rates, has performed the role of nominal anchor of the economy. Furthermore, results suggest that monetary policy has been consistent with inflation-targeting principles. This implies a central bank that immediately fights inflationary pressures stemming from the demand side and that lets cost-push shocks have a once-and-for-all effect on the price level. Interest rates in Mexico appear to be determined more in a forward- than in a backwardlooking manner, with interest rates responding to expected inflation rather than lagged inflation. Since the forward-looking component of monetary policy is a key element of an inflation-targeting framework, this result is consistent with the fact that the Bank of Mexico has formally adopted an inflation-targeting framework.

378

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

The best approximation to the process through which interest rates are determined is an augmented monetary policy rule that includes the expected inflation deviation from its target, the expected output gap, the country-risk perception and the difference between longand short-run domestic interest rates. The first two elements help reflect the relationship between interest rates and the fundamental performance of the economy. The third and fourth elements, respectively, capture the effect of uncertainty generated by international and domestic events on interest rates in Mexico. The evidence suggests that monetary policy in Mexico has contributed to inflation reduction in recent years. This means that fixed exchange-rate regimes, with endogenous monetary policy, are not the only option for a small open economy and in particular for emerging markets with high inflation. Independent monetary policy can be of help in reducing inflation, particularly if inflation-targeting frameworks serve to impose limits on the discretion of monetary policy. This means that even in the case of emerging-market economies under flexible exchange-rate regimes, monetary policy can become the nominal anchor of the economy.

Acknowledgements The author thanks Sven Arndt, Armando Baqueiro, Alejandro D´ıaz de León, Miguel Messmacher, Francisco Rodr´ıguez, Jéssica Roldán, Julio Santaella, Alejandro Werner and two anonymous referees for helpful comments and suggestions, as well as Miguel D´ıaz and Daniel Sámano for excellent research assistance. Opinions expressed in the article belong solely to the author and do not necessarily represent those of Banco de México.

References Ball, L. (1999). Policy rules for open economies. In J. Taylor (Ed.), Monetary policy rules. Chicago, IL: University of Chicago Press. Banco de México. (2002). Informe Anual. Various editions, Mexico. Barro, R., & Gordon, D. (1983). Rules, discretion and reputation in a model of monetary policy. Journal of Monetary Economics, 12(1), 101–121. Bernanke, B., Laubach, T., Mishkin, F., & Posen, A. (1999). Inflation targeting: Lessons from the international experience. Princeton, NJ: Princeton University Press. Clarida, R., Gal´ı, J., & Gertler, M. (1998). Monetary policy rules in practice. Some international evidence. European Economic Review, 42(6), 1033–1067. Clarida, R., Gal´ı, J., & Gertler, M. (1999, December). The science of monetary policy: A new Keynesian perspective. Journal of Economic Literature, 37(4), 1661–1707. Clarida, R., Gal´ı, J., & Gertler, M. (2000, February). Monetary policy rules and macroeconomic stability: Evidence and some theory. The Quarterly Journal of Economics, 115(1), 147–180. Clarida, R., Gal´ı, J., & Gertler, M. (2001, May). Optimal monetary policy in open versus closed economies: An integrated approach. American Economic Review Papers and Proceedings, 91(2), 248–252. Infosel. (2002). Encuesta Semanal sobre las Expectativas de los Especialistas en Econom´ıa del Sector Privado. Various editions, Mexico. Mart´ınez, L., Sánchez, O., & Werner, A. (2001). Consideraciones sobre la Conducción de la Pol´ıtica Monetaria y el Mecanismo de Transmisión en México. Documento de Investigación #2001–2, Banco de México.

A. Torres / North American Journal of Economics and Finance 14 (2003) 357–379

379

Kydland, F., & Prescott, E. (1977). Rules rather than discretion: The inconsistency of optimal plans. Journal of Political Economy, 85(3), 473–492. Rogoff, K. (1985, November). The optimal degree of commitment to an intermediate monetary target. Quarterly Journal of Economics, 100(4), 1169–1189. Rodr´ıguez, F. (2001). Identificación de una Regla de Pol´ıtica Monetaria Impl´ıcita en el Esquema de Objetivos de Inflación de México. Thesis ITAM. Schwartz, M., & Torres, A. (2002). Inflation expectations, country risk and monetary policy in Mexico. In Stabilization and monetary policy: The international experience. Mexico: Banco de México. Svensson, L. (1996). Inflation forecast targeting: Implementing and monitoring inflation targets. NBER Working Paper No. 5797. Svensson, L. (2000, February). Open economy inflation targeting. Journal of International Economics, 50(1), 155–183. Taylor, J. (1993). Discretion versus policy rules in practice. Carnegie-Rochester Series on Public Policy, North-Holland, 39. 195–214. Taylor, J. (1999). A historical analysis of monetary policy rules. In J. Taylor (Ed.), Monetary policy rules. Chicago: The University of Chicago Press. Woodford, M. (2001). The Taylor rule and optimal monetary policy. American Economic Review, 91(2), 232–237.