May 13, 2005 - economic discipline, inflation targeting is being promoted as a general .... GDP growth at 2.6% per year, well below the historical rates of the ...
Inflation targeting in Mexico: An empirical appraisal
Luis Miguel Galindo Jaime Ros May 2005 DRAFT
Paper prepared for the Amherst/CEDES Conference on Inflation targeting, Buenos Aires, May 13-14, 2005.
2
Inflation targeting (IT) has become increasingly popular over the past decade. As a nominal anchor for monetary policy with a public and explicit commitment to maintain economic discipline, inflation targeting is being promoted as a general framework in order to reduce and control the inflation rate, improve inflation predictability, accountability and transparency, reduce the expected inflation variability (Sheridan, 2001), improve the output-inflation trade off (Clifton et al, 2001), help solve the dynamic consistency problem and even reduce output variability (Svensson, 1997 and 1998). Several Latin-American countries such as Chile (1990), Peru (1994), Mexico (1999), Brazil (1999) and Colombia (1999) have moved their monetary regimes to an IT framework (Corbo, Landerretche, Schmidt, 2002 and Schmidt-Hebbel and Werner, 2002).
There are also a number of concerns regarding the adoption of an IT regime. There is an important concern about the ability of the Central Bank to control the inflation rate in particular under the presence of fiscal or external shocks. There are also fundamental doubts about the economic consequences of an IT regime. For example, Ball and Sheridan (2003) argue that the reduction of inflation and the reduction in output variability in the OECD countries is a general trend that is not necessarily related to the instrumentation of IT and that this kind of regime does not affect output variability. Furthermore, Newman and von Hagen (2002) claim that the evaluation of IT has been misleading because it has not properly considered the problem of regression to the mean given that some IT countries were worse off before their implementation than the countries without an IT framework.
In the case of emerging market economies there is a special concern about the relationship of IT with exchange rate movements, imperfect and poorly regulated financial markets and weak monetary institutions. In effect, there are important concerns about the effectiveness of an IT regime under weak fiscal conditions, poorly regulated financial systems, large potential external shocks, low institutional credibility and currency substitution phenomena (Fraga, et. al., 2003 and Calvo and Mishkin, 2003).
3 There is also a major concern that important shifts in exchange rates will affect the general competitiveness of the economy, the current account deficit and generate external shocks on the inflation rate. For example, abrupt changes in the exchange rate might lead to inflation paths that are inconsistent with the original inflation target and therefore the Central Bank might try to use the nominal exchange rate as a nominal anchor (Svensson, 1998). In this case, initial exchange rate shocks to the inflation rate are followed by a general appreciation of the real exchange rate affecting the general performance of the economy. Goldfajn and Gupta (2003) argue also that an appreciation of the real exchange rate is related with higher nominal interest rates or a tight monetary policy that makes economic recovery more difficult.
This paper addresses Mexico’s experience with inflation targeting. Section 1 presents some background on the operation of monetary policy in Mexico since the 1994-95 peso crisis. Section 2 assesses empirically a number of important issues in the evaluation of IT: the effects of the real exchange rate on output, the question of the pass through of exchange rate movements into inflation and the asymmetric response of monetary policy in the face of exchange rate shocks. Section 3 considers alternatives to IT as currently implemented. Section 4 concludes.
1. Inflation targeting in Mexico: some background
In the 1990s, Mexico has experienced a variety of monetary and exchange rate regimes. More precisely, the monetary regime has gone through three stages: nominal exchange rate targeting with a crawling band regime before the 1994 crisis, monetary targeting for a short period after the crisis, followed by a transition to inflation targeting that by now has been largely completed. With the 1994 crisis the exchange rate regime shifted from a crawling band to floating.
The experience with the crawling band regime and its crisis has been analyzed widely (see, for example, Lustig and Ros, 1998, and Ros, 2001). It was during this period, in 1993, that the Central Bank was given independence and a mandate to preserve price
4 stability. After the 1994-95 peso crisis, monetary policy focused on the growth of monetary aggregates and limits to credit growth as a means to control the inflationary effects of the sharp peso devaluations and rebuild the damaged credibility of the Banco de México. More precisely, the central bank established as its nominal anchor an intermediate target for the growth of the monetary base. As the country moved to a flexible exchange rate regime, an assumption of no international reserves accumulation was made. The ceiling on the growth of the monetary base was thus in essence a growth ceiling on net domestic credit. This monetary policy framework was soon abandoned as the policy failed to stabilize inflationary expectations, the exchange rate and inflation itself. The main reasons for this failure were the instability of the relationship between the monetary base and inflation and the fact that the central bank has no control of the monetary base in the short run. The demand for bills and coins in circulation largely determines the monetary base and this demand is very interest inelastic in the short term (Carstens and Werner, 1999, p. 14).
The peso crisis had strong inflationary effects, taking the inflation rate from single digits to over 50 percent. These inflationary effects of the sharp exchange rate depreciation were largely brought under control by means of a tight fiscal policy and an incomes policy negotiated with unions and business confederations. After this, the main objective of monetary policy became the reduction of inflation in a gradual and sustainable way. At the same time, the monetary policy regime moved towards influencing the level of interest rates establishing borrowed reserves as its key instrument with the monetary base becoming less relevant and the inflation target more important in the conduct of policy.
How does monetary policy operate?
Under this framework the Banco de México establishes that banks should seek a null balance in their accounts at the central bank. A penalty rate (double the CETES rate) is applied to overdrafts and positive balances are not remunerated. Thus, the banks seek this balance in particular because if the daily average balance were negative they would have to pay the penalty rate. By providing more or less liquidity, the overall net daily average
5 balance of all current accounts held by banks at the Banco de México may close the day at a predetermined amount. If negative, the central bank puts the banking system in “short” (“corto”) in which case at least one credit institution has to pay the penalty interest rate. This is the way in which the central bank exerts pressure on interest rates. Although the corto represents only a small amount of the total liquidity, increasing it induces banks to bid up interest rates to avoid paying overdraft charges and puts upward pressure on market interest rates. The reverse mechanism applies when the central bank targets a positive balance (“largo”) (Carstens and Werner, 1999, pp. 15-16; OECD, 2002, 2004).
The transition to inflation targeting was accelerated in January 1999 when the Banco de México announced a medium term inflation objective, and since 2000 the Central Bank publishes quarterly inflation reports to monitor the inflationary process, analyze inflation prospects and discuss the conduct of monetary policy. By now Mexico is considered to have in place the main components of an inflation targeting framework: an independent monetary authority (since 1993) that has inflation as its only policy objective, a flexible exchange rate regime, the absence of other nominal anchors and a “transparent” framework for the implementation of monetary policy (Schmidt-Hebbel and Werner, 2002, p. 37).
The implementation involves the choice of an inflation index to be adopted as target, the target range and the time horizon. The national CPI is used to determine the inflation objective. Since the beginning of 2000 the bank also tracks a core CPI which excludes volatile items — such as agricultural and livestock products, and also education (tuition fees) — and prices controlled by or agreed with the public sector. In the transition from high to moderate inflation rates, the inflation objective was specified in terms of a value that should not be exceeded. Today, policy specifies a target range of plus or minus a percentage point. From January 1999 onwards the target of monetary policy has been framed in terms of a medium term inflation objective of bringing down inflation to the rate prevailing in Mexico’s main trading partners by the end of 2003, interpreted as an annual growth of the CPI of 3 percent.
6
What has been the record of inflation targeting?
Table 1 shows the inflation targets since the 1994-1995 crisis, actual inflation performance, the growth projections and performance as well as the evolution of the real exchange rate. After a rough start when in 1995 the inflation objective was missed by over 30 percentage points, the central bank’s record has been improving over time with its inflation target being met in 4 out of the past 6 years and inflation tending to converge towards its medium term target range of 3 percent plus or minus one percentage point.
Table 1. Inflation, growth and the real exchange rate
Inflation target (%) Actual inflation (%) GDP growth projection
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
19
10
15
12
13
10
6.5
4.5
3±1
3±1
52.0
27.7
15.7
18.6
12.3
9.0
4.4
5.7
4.0
5.2 (p)
n.a.
>3
>4
5
3
4.5
n.a.
1.5
3.0
3.0 to
Actual -6.2 GDP growth Real exchange rate (end of the 141.4 year) (index)
3.5
5.1
6.8
4.9
3.7
6.6
127.8
114.1
116.1
105.8
100.0
-0.1
92.5
0.7
1.3
102.3
106.2
4.2 (p)
103.4
Source: Banco de México. The real exchange rate is defined as the nominal exchange rate to the United States dollar, multiplied by the ratio of foreign (consumers’ price index of United States) to domestic price levels (consumers’ price index of Mexico).
7 Along with the success in bringing down inflation, the growth performance has been disappointing. More precisely, in the second half of the 1990s, under the stimulus of a very competitive exchange rate, the economy rapidly recovered from the 1995 recession and grew at rates that often surpassed the growth rate projected by the central bank and the government. However, from 2001 to 2003, growth sharply decelerated to rates that for three years in a row were below or barely above the rate of population growth. That is, the economy recorded a decline in per capita incomes from 2000 to 2003 before recovering in 2004. Overall performance from 1994 to 2004 has been unsatisfactory with GDP growth at 2.6% per year, well below the historical rates of the period 1940 to 1980 (6 to 6.5 percent).
2. The empirical evidence
In this section we address three relevant questions for the evaluation of the IT regime. First, what are the short term and long run effects of the real exchange rate on output? Have these effects changed with trade liberalization and integration with the United States economy? Second, how important is the pass through of the exchange rate to prices? Has IT modified the pass through? Third, does IT have a bias towards exchange rate appreciation? If so, and the real exchange rate has long run effects on output, the regime is not neutral with respect to economic growth.
Empirical strategy
Our empirical analysis includes several econometric exercises. These econometric exercises try to assess the relevance of the real exchange rate effects on output, the passtrough of the exchange rate on inflation, and the possible asymmetry in the response of monetary policy to exchange rate shocks. However, IT monetary polices are relatively recent making difficult to use several econometric techniques. In this sense, these results must be considered only preliminary. The database consists of quarterly data for the period 1980 to 2003 and 1999 to 2003. The selection period was due to data limitations. A definition of the variables is included in the appendix.
8
The real exchange rate and output
The net effect of the real exchange rate on output is not clear-cut because there are alternative transmission channels with opposite effects in the short and long runs1. In order to assess the long run impact of the real exchange rate on output we proceed to estimate a vector autoregressive model (VAR) model including output, investment, United States output and the real exchange rate considering the order of integration of the variables and therefore testing for possible cointegration2 among them.
Table 2. Unit Root Test Variable
yt yt invt invt xyt xyt pt pt gapyt gapyt st st srt srt rt rt m2t m2t wt wt 1
ADF A -3.20(8) -5.41(5) -2.97(2) -5.62(2) -2.72(2) -5.16(1) -3.50(0) -8.81(1) -3.970 (8) -4.727 (8) -1.62(2) -4.75(1) -2.63(0) -10.76(0) -2.74(0) -8.73(1) -1.78(4) -2.79(4) -1.92(1) -3.02(4)
B 0.73(6) -5.23(5) -0.83(0) -5.54(2) -0.51(2) -5.19(1) -2.46(0) -8.83(1) -3.983 (8) -4.763 (8) -3.07(2) -3.88(1) -2.29(0) -10.77(0) -1.15(0) -8.53(1) -2.52(4) -1.71(4) 1.75(0) -2.64(4)
PP(5) C 2.92(6) -2.36(7) 0.59(0) -5.55(2) 4.31(2) -1.63(0) 1.73(0) -8.87(1) -4.027 (8) -7.781 (8) -1.57(3) -2.27(2) -0.02(0) -10.8(0) -0.72(0) -8.53(1) 0.63(4) -1.02(3) 4.66(0) -1.48(4)
A -4.97 -2.39 -2.43 -8.82 -2.79 -7.54 -0.38 -3.55 -8.714 -23.473 0.94 -8.62 -2.79 -10.70 -2.68 -9.88 -0.29 -7.30 -1.94 -12.02
B -0.21 -21.41 -1.13 -8.82 0.23 -7.61 -3.65 -2.50 -8.754 -23.501 -2.99 -7.83 -2.51 -10.71 -1.09 -9.72 -3.90 -5.55 2.14 -11.04
KPSS(10) C 2.77 -18.62 0.50 -8.84 7.02 -4.52 0.80 -1.65 -8.795 -23.501 -1.48 -6.38 -0.02 -10.76 -0.72 -9.75 5.46 -2.44 4.96 -9.45
μ 0.092 0.1701 0.913 0.0471 0.661 0.147 0.152 0.067 0.936 0.065 0.077 0.064 0.903 0.238 0.601 0.074 0.043 0.043 0.052 0.048 0.868 0.228 0.493 0.067 0.261 0.101 0.112 0.071 0.236 0.077 0.261 0.101 0.939 0.239 0.606 0.078 0.925 0.232 0.499 0.083
The literature on the contractionary effects of devaluation is very large, starting with the article by Krugman and Taylor (1978). On the empirical evidence for Mexico, see Kamin and Rogers (1997), Lopez and Guerrero (1998), and Kamin and Klau (1998). 2 An econometric analysis of the IT regime should include the potential presence of structural breaks in the series that can lead to spurious rejections in the order of integration and cointegration tests (Leybourne and Newbold, 2003). Hence, we also test for possible structural breaks in the cointegration framework.
9 Notes: Test statistics in bold indicate a rejection of the null hypothesis. Critical values at 5% significance level for the Augmented Dickey-Fuller and Phillips-Perron tests for a size T=100 are –3.45 including constant and trend (model A), -2.89 including constant (model B) and –1.95 without constant and trend (model C) (Maddala and Kim, 1998, p. 64). μ and is the KPSS test for the null hypothesis of stationarity around a level and deterministic linear trend, respectively. Both tests are calculated with a lag window size equal to 5. The 5% critical values for the two tests are 0.463 and 0.146, respectively (Kwiatkowski et. al. 1992, p. 166). Small letters represent the values in logarithms
The unit root tests (Dickey Fuller, 1981 and Phillips Perron, 1988) are included in Table 2. The main results of these tests indicate that all these variables are non stationary.3 Therefore, the VAR model including output (Yt), investment (INVt), United States output (YXt) and the real exchange rate4 (SRt) should be analyzed considering the possible presence of a cointegrating vector and possible presence of structural breaks. The statistics of the Johansen (1988) procedure (Table 3) indicates the presence of at least one cointegrating vector (lower case letters refer to the natural logarithm of the series).
Table 3. Statistics of the Johansen procedure including output, investment, United States output and the real exchange rate. Period 1981:01-2003:04 yt = 1*it + 2*yust + 3*srt
H0 r=0 r1 r2 r3
Constant m0 m0 m0 m0
Trend 0 0 0 0
Trace 63.09* 20.57 6.10 0.01
95% 47.21 29.68 15.41 3.76
Notes: (*) Significant at the 5% level; Trace = Trace test; r= number of co-integrating vectors. Number of lags in the VAR = 4; The VAR includes constant unrestricted.
Normalizing this vector as an output equation (equation 1) indicates the presence of a positive relationship between output, investment, United States output and the real
3
However, it is also possible to argue that the real exchange rate is stationary in the long run considering structural changes. 4 The real exchange rate is defined as: SRt = St(Pus/P)t, where S is the nominal exchange rate; Pus, the consumer price index of the United States: and P, the consumer price index of Mexico.
10 exchange rate. Therefore, a devaluation of the real exchange rate has a positive impact on Mexican economic growth in the long run.
(1)
yt = 0.478*invt + 0.320*yxt + 0.019*srt
The analysis of the Johansen procedure designed to address the possibility of structural breaks in the equilibrium relationship still indicates the presence of a long-term relationship among the variables (see Figures 1.a and 2.a in the Appendix). The impulse response analysis, using the Cholesky decomposition and the same order of the variables as in the Johansen procedure, confirms the positive effect of investment and United States output on Mexico’s economic growth and indicates that the real exchange rate has an initial negative effect that tends to disappear in the long run (Figure 1). Additionally, the variance decomposition analysis indicates that the real exchange rate is certainly relevant in order to explain output movements (Table 1.a in Appendix).
11 Figure. 1. Impulse – Response analysis for output, investment, United States output and the real exchange rate
The finding that an appreciation of the real exchange rate has long run contractionary effects on economic growth in Mexico contradicts previous results suggesting that the long run relationship between output and the real exchange rate is negative (Kamin and Rogers, 1997). It is also worth noting that the effect of the real exchange rate on output has significantly increased after the establishment of the North American Free Trade Agreement (Table 3 and Table 2.a in the Appendix). Therefore, the Mexican economy seems to be more sensitive to exchange rate changes today than in the past.
12 Table 3. Cointegration vectors of the Johansen procedure including output, investment, United States output and the real exchange rate. yt = 1*it + 2*yust + 3*srt
Period
invt
yxt
srt
1982(1)-1993(4) 1994(1)-2003(4)
0.852 0.761
0.328 0.420
0.475 0.855
The nominal exchange rate and inflation (Pass trough)
The positive impact of the nominal exchange rate on the inflation rate (pass trough) is one of the main concerns of the monetary authorities. Under these conditions, an IT regime is prone to suffer from external dominance. That is, the high sensitivity of inflation to any exchange rate depreciation could lead to external inflation shocks that will make it difficult for the central bank to achieve its inflation target. On the other hand, it has also been argued that an IT regime reduces the pass-trough problem due its ability to increase the credibility of the monetary authorities and therefore the importance of forward-looking variables (Schmidt-Hebbel and Werner, 2002 and Fraga, Goldfajn and Minella, 2003).
We evaluate the pass trough effect in Mexico using econometric tests. First, we estimate a VAR including the price index (pt), the output gap (gapyt), the nominal wage (wt), the nominal exchange rate (st) and a monetary aggregate (M2) (mt)5. The sample period covers quarterly data from 1986:01 to 2003:04. This VAR includes variables with different order of integration.6 For example, the output gap is I(0) and arguably the price
5 6
Cartens and Werner (1999) include additionally the public price index and use base money instead of M2. See Table 2
13 index is I(2).7 Therefore, we use the impulse response and the variance decomposition analysis because the cointegrating vector might impose unreliable restrictions. That is, the Johansen (1988) procedure (Table 3.a in the Appendix) indicates the presence of at least one cointegrating vector normalizing as a price equation (equation 1.a in the Appendix). In this case, the output gap, wages, money and the nominal exchange rate have a positive impact on the price index. However, the sum of all the long term coefficients is larger than one. Arguably, the high level of these coefficients is related to the inclusion of variables with different order of integration and furthermore, the causality is not clear-cut. For example Carstens and Werner (1999) find that base money is essentially accommodating to exogenous shocks.
The impulse response analysis also confirms the positive impacts of these variables on the price index. However, it also indicates that the effect of the exchange rate and the monetary aggregate is more robust than the other variables (Figure 2). Additionally, the variance decomposition in the Appendix (Tables 4.a,) confirms the relevance of the exchange rate movements in order to explain the variance of the price index. These results clearly confirm the relevance of the pass trough effect from the nominal exchange rate to the price index in Mexico.
7
See unit root test in Table 2.
14
Figure 2. Impulse – Response Analysis for the price index, output gap, wages, nominal exchange rate and a monetary aggregate
Second, we estimate a VAR including the inflation rate, the output gap and changes in the nominal exchange rate. The sample period covers quarterly data from 1986:01 to 2003:04. This specification is more similar to a traditional Phillips curve and includes some elements of the VAR specification used in Schmidt-Hebbel and Werner (2002) or Fraga, Goldfajn and Minella (2003). Additionally, it is possible to argue that this VAR includes variables that are all I(0) (Table 2). The impulse response (Figure 3) and the variance decomposition (Table 5.a) indicate that the output gap and the change in the exchange rate have both a positive impact on the inflation rate also reflecting the relevance of the pass trough effect.
15
Figure 3. Impulse – Response Analysis for the inflation rate, the output gap and the changes in the nominal exchange rate
In order to evaluate the relevance of the pass trough in the Mexican economy we can consider the “rolling” correlation coefficient between the inflation rate and the exchange rate depreciation. This correlation coefficient is estimated by adding one observation sequentially since 1989 (I). Figure 4 indicates the presence of a strong relationship between these two variables. The initial reduction of the pass trough, arguably related with the instrumentation of the IT regime, is not continuous and there is a persistence of a positive relationship between inflation and exchange rate movements.
16
Figure 4. Rolling correlation coefficient between the inflation rate and the exchange rate depreciation Quarterly data 1989:1-2003:4
The response of monetary policy to exchange rate shocks
In this section, we look for asymmetric effects of the exchange rate on monetary policy. The main hypothesis is that monetary policy shows an asymmetric response to movements in the exchange rate. That is, the central bank raises the interest rate in response to a depreciating exchange rate but does not modify the interest rate in response to an appreciating exchange rate. The final result of this monetary policy is an appreciation of the real exchange rate.
Table 3 shows the changes in the “corto” that have taken place between January 1996 and December 2004 and figure 5 shows these changes together with the monthly evolution of the peso-dollar exchange rate. Over this period, the Banco de Mexico has increased the “corto” on 34 occasions and reduced it six times. Never over the period has the banking
17 system been put in a “largo”. Increases in the “corto” were preceded by or were simultaneous to depreciations of the peso on 18 occasions (53% of the number of increases) and the increase in the corto was followed by appreciations of the peso on 14 occasions. Only on four occasions, August and November 1996 and May and July 2001), were appreciations followed by a reduction of the “corto” and only on one occasion did the Central Bank allow a sustained depreciation of the peso to take place without increasing the “corto” (from June through December of 2003).
Table 3. Changes in the “corto”, 1996-2004 (million pesos)
Date 1996 January 23
Changes in the “corto” Increase from 0 to 5
January 25 June 7 June 21 August 5
Increase from 5 to 20 Increase from 20 to 30 Increase from 30 to 40 Reduction from 40 to 30
August 19
Reduction from 30 to 0
October 14
Increase from 0 to 20
November 26 1998 March 11 June 25 August 10 August 17 September 10 November 30 1999 January 13 2000 January 18 May 16 June 26 July 31 October 17 November 10
Reduction from 20 to 0 Increase from 0 to 20 Increase from 20 to 30 Increase from 30 to 50 Increase from 50 to 70 Increase from 70 to 100 Increase from 100 to 130 Increase from 130 to 160 Increase from 160 to 180 Increase from 180 to 200 Increase from 200 to 230 Increase from 230 to 280 Increase from 280 to 310 Increase from 310 to 350
18 2001 January 12
Increase from 350 to 400
May 18 July 31 2002 February 8 April 12 End of September
Reduction from 400 to 350 Reduction from 350 to 300
December 6 2003 January 10
Increase from 400 to 475
February 7 March 28 2004 February 20 March 12 April 27 July 23 August 27 September 24 October 22 November 26 December 10
Increase from 300 to 360 Reduction from 360 to 300 Increase from 300 to 400
Increase from 475 to 550 Increase from 550 to 625 Increase from 625 to 700 Increase from 25 to 29 (daily) Increase from 29 to 33 Increase from 33 to 37 Increase from 37 to 41 Increase from 41 to 45 Increase from 45 to 51 Increase from 51 to 57 Increase from 57 to 63 Increase from 63 to 69
Source: Bank of Mexico, Informe Anual and Informe sobre Politica Monetaria, various issues.
We also explore the possibility of an asymmetrical response of monetary policy to exchange rate movements with an econometric exercise. In order to assess the relevance of this asymmetric effect we use a two-step procedure similar to Cover (1992), Karras (1996) and Kim, Ni and Ratti (1998). This procedure contains initially an equation in order to obtain an equilibrium exchange rate trough the use of the PPP hypothesis (Hallwood and MacDonald, 2000). We consider that a value above the equilibrium position is an undervaluation while a value under the equilibrium position is an overvaluation. These values are, afterwards, used to test for a possible asymmetric response of monetary policy to movements in the exchange rate.
19 Hence, the first equation describes the exchange rate using the purchasing power parity condition (Hallwood and MacDonald, 2000):
(1)
Where
denotes the nominal exchange rate,
denotes the national price index and
denotes the foreign price index.
Information for such variables comes from Mexico’s central bank database. The data base includes quarterly information from 1995(1) to 2004(4) on the nominal exchange rate and the price indexes of Mexico and United States. The estimation of equation (1) using ordinary least squares (OLS) indicates that the nominal exchange rate has a longterm relationship with the price index differentials.
The estimated long run equation is as follows: (2)
= 3.36+ 1.202*
ADF(1) = -3.07 Then, using equation (2) we define two shock functions based on the differences between the fitted values and the actual values of the exchange rate (U). Pos U+t = max (shock, zero) Neg U-t = min (shock, zero) The first function is associated to an undervaluation and the second one refers to an overvaluation. The second step of the procedure consists in estimating an interest rate equation including the positive and negative shocks given by the previous functions.
(3)
20
Equation (3) describes the impact of exchange rate deviations on the interest rate. This equation is estimated to find evidence of asymmetric effects of the exchange rate on the interest rate. The interest rate is the three months nominal interest rate on government bonds (CETES) (Mexico’s Central Bank data base). Making use of such a procedure we get the following estimates:
=
(3) (t) (p)
-0.087*
+ 0.145*
+ 0.76*
(-1.01) (0.26)
(3.16) (0.00)
(25.70) (0.00)
- 23.6*D98(3) – 20.7*D95(4) (-7.95) (0.00)
(-6.72) (0.00)
R2=0.82 Normality test8: Jarque-Bera: X2(2) = 2.11(0.35) Autocorrelation: LM (4) = 0.169(0.952) Heteroskedasticity: ARCH (4) = 1.211(0.326) Period: 1995:1-2004:4
The long run solution is: (4)
t
= -0.371*
+ 0.615*
The main results indicate that only in the case of an undervalued exchange rate we find a statistically significant coefficient. According to such estimates a tight monetary policy is carried out by the central bank whenever there is an undervalued exchange rate; however an overvalued exchange rate is not followed by an easy money monetary policy. The previous estimates thus confirm that the exchange rate has an asymmetric impact on monetary policy.
8
The non-normality can be attributed to an outlying value in 1995:2 when debt and financial crisis took place in Mexico.
21 3. Problems with inflation targeting and alternatives
As discussed in section 1, the recent past has been characterized by success in the inflation front and a poor growth performance. Two of our findings in section 2, the positive effect of the real exchange rate on output and the significant pass through of the exchange rate movements on prices, suggest that those two aspects, success in the inflation front and poor growth performance, are linked through the evolution of the real exchange rate (see table 1). After the sharp depreciation of 1995, the real exchange rate has been appreciating over time, a trend that was only interrupted in 2002-03. From end of 1995 to 2002, the real exchange rate fell by 35 percent (and by 27 percent from 1995 to 2004). With a relatively stable nominal exchange rate, in an increasingly open economy, government policy has provided a strong disinflationary pressure. At the same time, real appreciation, through the loss of competitiveness of the economy, has contributed to a poor growth performance.
Our third finding in section 2, on the asymmetric response of monetary policy to exchange rate shocks, indicates that real appreciation may be the result of a built in bias in monetary policy towards real exchange rate appreciation. This bias has to do with the high pass through. Over the past few years the central bank has been trying to break the link between exchange rate and prices by increasing the “short” in response to “sharp” depreciations. In doing so, it has reversed the depreciation itself. Because the economy has been in a process of disinflation in which the central bank has barely met its inflation targets, the process is not symmetrical, i.e., there is no similar incentive to reverse the appreciations that may take place as a result of shocks to the exchange rate.
Alternatives
What are the alternatives?
A first option is to move monetary policy towards a more neutral stance (i.e. a more symmetric response to exchange rate shocks). This is more likely and feasible than in the
22 past as inflation tends to converge towards the long run target and, as a result of either an enhanced credibility of the central bank or a less inflationary environment, the pass through of the exchange rate on inflation tends to fall.
A second option is to shift from a CPI target to a domestic inflation target (i.e. a measure of inflation purged from the direct effects of the exchange rate on imported goods in the CPI). Such a move would further reduce the pass through effect of the exchange rate on the targeted price index and contribute to eliminate the asymmetric response of monetary policy to exchange rate shocks.
These two options preserve the inflation targeting framework. A third, more radical departure from the current framework, would be to combine inflation targeting with real exchange rate targeting. More precisely, the central bank would promote a competitive exchange rate by establishing a sliding floor to the exchange rate in order to prevent excessive appreciation (an “asymmetric band” as in Ros, 1995). This would imply intervening in the foreign exchange market at times when the exchange rate hits the floor but allow the exchange rate to float freely otherwise.
The orthodox objection to such a proposal is that by defending the floor the central bank loses control of the money supply and this could imply giving up the achievement of the inflation target (for a fuller discussion see Frenkel and Rapetti, 2004). The problem arises at times of excess supply of foreign currency as a result, in particular, of massive capital inflows. It is worth noting, however, that speculative capital inflows will tend to be deterred to the extent that the central bank clearly signals that it will prevent the appreciation of the domestic currency thus stabilizing exchange rate expectations. If necessary, however, the central bank can impose capital account regulations on shortterm capital flows in order to recover control over the money supply.
23
4. Conclusions
The main conclusions that emerge from our analysis can be summarized as follows. Inflation targeting has contributed to a gradual reduction in inflation from levels over 50% in 1995 to around 5% in 2004. It may also have contributed to a reduction of the pass through of exchange rate movements into inflation possibly as a result of an enhanced central bank credibility (with its stabilizing influence on inflation expectations) or of lower inflation itself. These achievements have, however, come with a cost, an almost continuous appreciation of the real exchange rate that has had contractionary effects on output in the long run. Our empirical analysis clearly supports these assertions (the reduction of the pass through and the negative effect of real appreciation on output). It also suggests that real appreciation has been fed by an asymmetrical response of monetary policy to exchange rate movements: depreciations are followed by a tightening of monetary policy while appreciations are not reversed by a relaxation of monetary conditions. The alternative to the current framework is to give a more prominent place to the achievement of a competitive and stable real exchange rate in the design of monetary and exchange rate policy.
References
Ball, L. and N. Sheridan (2003), “Does inflation targeting matters?”, Working Paper No. 9577, NBER Working Paper Series, March. Calvo, G. A. and F.S. Mishkin (2003), "The Mirage of Exchange Rate Regimes for Emerging Market Countries", Journal of Economic Perspectives, 17(4) Cartens, A. and A. Werner (1999), “Mexico´s monetary policy framework under a floating exchange rate regime”, Documento de Investigación No. 90-05, Banco de México, mayo.
24 Castellanos Sara G. (2000), “El efecto del “corto” sobre la estructura de las tasas de interés”, Documento de Investigación No. 2000-1, Banco de México, Junio. Clifton, E. V., H. Leon and C.H. Wong (2001), “Inflation targeting and the unemployment-inflation trade off”, WP/01/166, Paper IMF Working Paper, International Monetary Fund (IMF). Corbo, V., O. Landerretche y K. Schmidt-Hebbel (2002), “Does Inflation Targeting Make a Difference?”, in Inflation Targeting: Design, Performance, Challenges , (eds) N. Loayza and R.Soto. Santiago: Banco Central de Chile. Cover, J.P. (1992), “Asymmetric Effects of Positive and Negative Money-Supply Shocks”, The Quarterly Journal of Economics, 107(4): 1261-1282. Dickey, D. and W.A. Fuller (1981), “Likelihood ratio statistics for autoregressive time series with unit root”, Econometrica, 49(4): 1057-1977. Fraga, A., I. Goldfajn and A. Minella (2003), Inflation targeting in emerging market economics, NBER Macroeconomics Annual. Frenkel, R. and M. Rapetti (2004), “Políticas macroeconómicas para el crecimiento y el empleo”, paper prepared for the ILO Goldfajn I. And P. Gupta (2003), “Does Monetary Policy Stabilize the Exchange Rate Following a Currency Crisis?”, IMF Staff Papers, 50(1), International Monetary Fund. Hallwood C. and R. MacDonald (2000), International money and finance, Blackwell publishers Johansen, S. (1988), “Statistical analysis of cointegrating vectors”, Journal of Economics Dynamics and Control, 12(2-3): 231-54. Kamin, S.B. and J.H. Rogers (1997), “Output and the real exchange rate in developing countries: an application to Mexico”, International Finance Discussion Papers, No. 580, Board of Governors of the Federal Reserve Bank, May. Kamin, S.B. and M. Klau (1998), “Some multi country evidence on the effects of real exchange rates on output”, International Finance Discussion Papers, No. 611, Board of Governors of the Federal Reserve Bank, May.
25 Karras, G. (1996), “Are the Output Effects of Monetary Policy Asymmetric? Evidence from a sample of European Countries”, Oxford Bulletin of Economics and Statistics, 58(2): 267-278. Kim, J., S. Ni and R. Ratti (1998) "Monetary Policy and Asymmetric Response in Default Risk", Economics Letters, 60(1): 83-90. Krugman, P. and L. Taylor. (1978), "Contractionary Effects of Devaluation", Journal of International Economics, 8(3): 445-456. Kwiatkowski, D., P.C.B. Phillips, P. Schmidt e Y. Shin (1992). "Testing the null hypothesis of stationary against the alternative of a unit root", Journal of Econometrics, 54(1-3): 159-178. Leybourne, S. and P. Newbold (2003) “Spurious rejections by cointegration tests induced by structural breaks”, Applied Economics, 35(9): 1117-1121. López, J. and C. Guerrero (1998), “Crisis externa y competitividad de la economía mexicana”, El Trimestre Económico, October-December, LXV(260): 582-598. Lustig N. and J. Ros (1998), “Reforma estructural, estabilización económica y el síndrome mexicano", Desarrollo Económico, 37(148): 503-531. Maddala, G.S. e I. Kim (1998). Unit roots, Cointegration and Structural Change, Cambridge University Press. Neumann, M.J.M. and J. von Hagen (2002), “Does inflation targeting matter?”, Federal Reserve Bank of St. Louis. Review, July-August, 2002. OECD (2002), Economic Surveys, Mexico, OECD OECD (2004), Economic Surveys, Mexico, OECD Phillips, P.C.P. and P. Perron (1988), “Testing for unit root in the time series regression”, Biometrica, 75(2): 335-346. Ros, J. (2001), “Del auge de capitales a la crisis financiera y más allá: México en los noventa”, in Ricardo Ffrench-Davis (ed.) Crisis financieras en países “exitosos”, CEPAL-McGraw Hill. (1995), “Mercados financieros, flujos de capital y tipo de cambio en México”, Economía Mexicana, CIDE, IV(1): 5-68 Sheridan, N. (2001), “Inflation Dynamics”. Johns Hopkins University. PhD. Dissertation.
26 Schmidt-Hebbel, K. and A. Werner (2002), “Inflation targeting in Brazil, Chile and Mexico: Performance, credibility, and the exchange rate”, Central Bank of Chile, Working Paper No. 171, July Svensson, L.E. (1998), “Open economy inflation targeting”, Working Paper No. 6545, NBER Working Paper Series, May. Svensson, L. E. (1997), “Inflation Forecast Targeting: Implementing and Monitoring Inflation Targets”, European Economic Review, 41(6): 1111-1146. APPENDIX Graph 1.a Stability of cointegration rank: The R-model
27 Graph 2.a Stability of cointegration rank: The Z-model
Table 1.a Variance Decomposition of yt (%)
Períod
S. E.
yt
it
yxt
srt
1 2 3 4 5 6 7 8 9 10
0.017343 0.024585 0.030544 0.035180 0.040243 0.043733 0.046420 0.048517 0.051032 0.052671
100.0000 63.88297 52.49081 43.07991 48.60616 44.60185 43.56448 41.36933 45.37348 44.84899
0.000000 26.16900 28.75609 30.25557 23.33574 25.69707 25.78104 27.45829 24.87671 25.97610
0.000000 1.836138 3.027916 3.370202 3.146353 3.413014 3.665154 3.917839 3.844143 4.018295
0.000000 8.111887 15.72518 23.29431 24.91175 26.28807 26.98933 27.25453 25.90567 25.15662
28 Table 2.a Johansen cointegration test yt = 0*it + 1*yxt + 2*srt
r=0 r1 r2 r3
1980:01-1993:04 Trace 95% 54.57* 39.9 24.3 27.75* 6.244 12.5 1.842 3.8
1994:01-2003:01 Trace 95% 41.75* 39.9 21.91* 24.3 10.48 12.5 1.349 3.8
Notes: * significant at the 5% level. Trace = trace test; r = number of cointegrating vectors. Number of lags in the VAR=5, the VARs including constant unrestricted.
Table 3.a Johansen cointegration tests pt = 1*gapyt + 2*wt + 3*st + 4*m2t
Ho
Trace
95%
85.9143* 68.5 r= 0 34.477 47.2 r 1 10.497 29.7 r 2 1.8877 15.4 r 3 0.5728 3.8 r 4 Notes: (*) Significant at the 5% level; Trace = trace test; r = number of co-integrating vectors. Number of lags in the VAR = 6; The VAR includes variables dummy 1987:04, 1988:02 and 1995:01 unrestricted; Period 1986:01-2003:04.
(1.a)
pt = 0.389*gapyt + 0.926*wt + 0.917*st + 0.543*m2t
29 Table 4.a Variance decomposition of pt (%) Period
S. E.
pt
Gapyt
wt
st
m2t
1 2 3 4 5 6 7 8 9 10
0.023307 0.048459 0.073098 0.100729 0.131896 0.165813 0.197532 0.227077 0.254025 0.278260
100.0000 89.42972 69.37043 48.46701 34.20141 24.53181 18.66775 14.91181 12.44775 10.71272
0.000000 0.491010 0.668013 1.400164 1.847610 1.793723 1.684104 1.556711 1.516772 1.528304
0.000000 0.301881 0.194002 0.113535 0.093912 0.104776 0.125739 0.177886 0.271083 0.319327
0.000000 8.745394 21.16145 32.23509 41.37765 47.49545 50.51930 52.93990 54.39805 54.85820
0.000000 1.032000 8.606107 17.78420 22.47941 26.07424 29.00310 30.41369 31.36634 32.58145
Table 5.a Variance Decomposition of pt (%) Period
S. E.
pt
gapyt
st
1 2 3 4 5 6 7 8 9 10
0.031432 0.045856 0.051234 0.055749 0.061895 0.067112 0.070685 0.074272 0.077747 0.080887
100.0000 96.77959 92.94494 89.11708 87.58702 86.75209 86.12336 84.67810 83.50346 82.11481
0.000000 0.100551 0.740052 3.429786 5.129741 6.729184 7.995716 9.958204 11.32048 12.46825
0.000000 3.119855 6.315006 7.453139 7.283235 6.518725 5.880923 5.363699 5.176066 5.416936
Database
Y = Real Gross Domestic Product (GDP) in millions of Mexican pesos of 1993, Instituto Nacional de Geografía, Estadística e Informática (INEGI), http://www.inegi.gob.mx.
30 YX = Real Gross Domestic Product (GDP) in billons of US dollars at prices of 2000, (the series is already seasonally adjusted), US Department of Commerce: Bureau of Economic Analysis. P = Mexican consumers’ price index (base 2002 = 100), Bank of Mexico, http://www.banxico.org.mx. P* = Consumer price index of US (base 1982-84 = 100), US Department of Commerce: Bureau of Economic Analysis. GAPY = Deviation of the real Gross Domestic Product (GDP) from potential output obtained using the Hodrick Prescott filter. S = Exchange rate (pesos per US dollar), 48-hour interbank exchange rate. The data is taken from the last day of each quarter, Bank of Mexico. http://www.banxico.org.mx. SR = Real exchange rate defined as sr = s (p*/p), where s is the logarithm of the nominal exchange rate, p the logarithm of domestic prices and p* the logarithm of foreign prices. R = Nominal interest rate on three-month treasury bills (CETES 91 days). Average of the last month of the quarter, Bank of Mexico, http://www.banxico.org.mx. M2 = Monetary aggregate (M2) in millons of Mexican pesos, Bank of Mexico, http://www.banxico.org.mx. W
=
General
minimun
http://www.banxico.org.mx.
wage,
pesos
per
day,
Bank
of
Mexico,
