5TH INTERNATIONAL WORKSHOP ON EUROPEAN ECONOMY

22 AND 23 NOVEMBER 2002

CEDIN / ISEG

THE PORTUGUESE DISINFLATION PROCESS:

ANALYSIS OF SOME COSTS AND BENEFITS

António Duarte

(http://www4.fe.uc.pt/portugal/ - [email protected])

FACULTY OF ECONOMICS – UNIVERSITY OF COIMBRA

GRUPO DE ESTUDOS MONETÁRIOS E FINANCEIROS

(GROUP FOR MONETARY AND FINANCIAL STUDIES – GEMF) Av. Dias da Silva, 165  3004-512 COIMBRA, PORTUGAL http://www4.fe.uc.pt/gemf/

I. Introduction

The present study main goal consists in analysing the Portuguese economic policy of disinflation through a nominal stabilization policy of the Portuguese Escudo.

We study the pegging of the Portuguese Escudo (PTE) to the Deutsch Mark (DM) taking for granted the anti- reputation of the Bundesbank and the role played by the Deutsch Mark in the stability process of the foreign exchange and of the European price level.

This presentation is structured in the following way:

Section II presents a historical retrospective of the disinflation process in the Portuguese economy and analyses some costs and benefits of this process.

Section III lay out of the data and handling of the series used.

Section IV analysis of the existence of a co-integration relation.

Section V describes the construction methodology of a Near-Var model for both Portuguese and German economies and it analyses also its main features.

Section VI complements the latter section, evaluating the results of a simulation analysis.

Finally, section VII concludes this work, leaving some clues for future research.

1 II. The Portuguese Disinflation Process: The Role of the Nominal Stabilization Policy of the Portuguese Escudo

The disinflation experience of the Portuguese economy, while a gradual and continued process in the reduction of the inflation rate, began in 1990 when Portugal adopted a nominal stabilization policy of the Escudo in view of other countries with a long tradition related to price stability, where the Deutsch Mark was seen as a reference .

Figure 1 shows the evolution of the Portuguese and German inflation rates in the periods between 1979 and 1998 and between 1990 and 1998.

Figure 1: Portuguese and German Inflation Rates

30 TxInfP TxInfA

20

10

0 1980 1985 1990 1995 2000 30 TxInfP TxInfA

20

10

5

0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Source: OECD data base Main Economic Indicators, published by Estima, USA, www.estima.com.

As it is easily observed, after the rise of the inflation rates in the beginning of the 80s, the Portuguese inflation rate reached a maximum value of about 30% by the end of 1983, recording, too, at that time the biggest differential compared to Germany.

Such a high rate inflation, reflected the successive devaluation’s of the Escudo exchange rate in an attempt to maintain the foreign competitiveness of the Portuguese industry.

After ten years, Portugal sees its value reduced to about 5%, and continuing this dropping route up until today.

This disinflation process is associated to the adoption, by the Portuguese monetary authorities, of a nominal stabilization policy of the Portuguese Escudo, in the context of its participation in the Exchange Rate Mechanism (ERM) of the (EMS), as it is illustrated in Figure 2. 2 Figure 2: Nominal variation of the Escudo to the Mark

∆ NomEM 20

10

0

1980 1985 1990 1995 2000

∆ NomEM 20

10

5 0

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Source: OECD data base Main Economic Indicators, published by Estima, USA, www.estima.com.

The exchange rate policy was therefore favouring the anti-inflation goal leaving behind consensus about reinforcing foreign competitiveness of the Portuguese economy.

In fact, in the period prior to Portugal adhesion to the European Economic Community (EEC), the exchange rate regime was characterised by the maintenance of a crawling peg regime of the Escudo.

However, the success of the integration in the European market was deeply dependent on the capacity of the monetary authorities to reduce the inflation rate to the levels presented by Germany.

In this context, in October 1990, the crawling peg policy was replaced by a policy based on a limited floatation of the Escudo in relation to the five main currencies of the ERM of the EMS. Portugal began, then, adopting an exchange rate policy based on the nominal stability of the Escudo.

Since the Portuguese inflation rate registered a high convergence rate to the European levels, the Portuguese government decided to propose the incorporation of the Escudo in the ERM of the EMS, becoming effective on the 6th April 1992. The discipline of this mechanism functioned as an anchor of the disinflation policy.

Nevertheless, in September 1992 we witnessed a deep crisis in the EMS translated into a succession of some realignments that culminated with the expansion of the foreign exchange floatation bands to fifteen % (15%) in August 1993.

In this context, the Escudo ended up by devaluating in 6% and 6,5%.

Nevertheless, the Portuguese Escudo realignments didn’t aim at changing the generic option for a foreign exchange stability policy that was being followed. 3 It is now important to analyse some of the costs and benefits associated with the disinflation process.

In fact, with the exception of the period between 1992 and 1994, the disinflation process was also characterized by a real appreciation of the Escudo to the Deutsch currency, as illustrated in Figure 3.

Figure 3: Real Variation of the Escudo to the Mark

∆ RealEM 10

0

-10

1980 1985 1990 1995 2000

∆ RealEM 10

5

0

-5

-10 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Source: OECD Data Base Main Economic Indicators, published by Estima, USA, www.estima.com.

In face of this evidence, a question a rises about how a small open economy like the Portuguese one, with a weak currency, didn’t experience a financial crisis due to the speculative attacks that were registered against the Escudo?

This was due to the fact that the Portuguese economy was involved in a process of European integration, allowing it to benefited from all the credibility, stability and discipline conferred by the tacit acceptance of the anti-inflation monetary policy of the Bundesbank and by the pegging of the Portuguese Escudo to the Deutsch Mark.

Due to this strategy, the Portuguese economy was also able to reach successfully the main goal of price stability, that was explicitly assumed by the political authorities.

Nevertheless, it should also be mentioned that as a result of the disinflation process there was a loss of Portuguese competitiveness due to the real appreciation of the Portuguese Escudo, limiting, thus, the product growth.

In this context, this study tries to clarify the influence that an appreciation of the real exchange rate can have on GDP and price level.

4 III. Data

Time series were used with quarterly data covering the period between 1977 and 1998.

In general, the data was obtained from the statistical sources published by OECD (Main Economic Indicators), being, preferably, used the most recent ones. For Portugal we also used data from the National Institute of Statistics and from the Bank of Portugal.

The price levels and output series were expressed in simple indexes based on the year of 1990. GDP was evaluated at constant prices. Short run nominal interest rates were used. With the exception for the interest rates, all series were used in logs. The majority of the results were obtained by using econometric software PcGive, version 10.

IV. Co-Integration

The co-integration analysis was developed with the aim of finding out if in the long run it is possible to identify a linear combination between domestic macroeconomic variables that is stationary.

The estimation technique used in the co-integration time series consisted in the application of the maximum likelihood method developed by Johansen.

Once identified the variables to be included in the system and once defined its nature, number of lags to be used were then chosen, in order to avoid errors auto-correlation problems.

We reached, therefore, eight lags for the endogenous variables, and no lags were retained for the exogenous variables.

Having as a base the values of the λ max (maximum eigenvalue) and the trace tests for a confidence interval of 95%, the number of the co-integration vectors found coincides with the number of the endogenous variables used, as it can be observed in Table 1.

Table 1: Co-integration Analysis (λ max and trace tests) λ r = p λ max λ max (A) 95% Trace Trace (A) 95% 0.458132 p = 0 49.02** 29.41* 27.1 120.8** 72.48** 47.2 0.410487 p ≤ 1 42.28** 25.37* 21.0 71.78** 43.07** 29.7 0.225008 p ≤ 2 20.39** 12.24 14.1 29.5** 17.7* 15.4 0.107651 p ≤ 3 9.112** 5.467* 3.8 9.112** 5.467* 3.8 (A) It was used the correction of freedom grade, the observations number (T) was replaced by T-n.m, where n is the number of variables and m is the number of lags.

In this situation, we may conclude that the variables are jointly stationary.

It were, therefore, established the conditions that allowed us to move forward to the construction and estimation of a Near-VAR model for both Portuguese and German economies.

5 V. A Near-VAR Model for the Portuguese and German Economies

In order to construct the Near-VAR model it was necessary to proceed, initially, to the formulation and estimation of a system that would serve as the base to all the construction process.

The system was always estimated through the Ordinary Least Squares (OLS) method.

Completed the first estimation, the process of model reduction begins, having as a base the significance level of the t statistic (t-prob value) of each of the variables. The aim is to eliminate from the formulation model the variables that present themselves as statistically insignificant.

We were now in presence of a Near-VAR model. Its estimation was done through the Full Information Maximum Likelihood (FIML) method.

Nine reductions were done. Thus, it was not possible to continue with the reduction process. All the variables that formed the equations model were significant. As it can be observed by the final results of model reduction tests, it were eliminated a total of seventy-eight (78) variables, because they were statistically accepted as insignificant:

Final results of the reduction tests model: Model 2 Æ Model 11: Chi^2 (78) = 73.101 [0.6357] System 1 Æ Model 11: Chi^2 (78) = 95.846 [0.0830]

Concerning the residuals analysis, it was necessary to find out if there were auto-correlation, normality and heteroscedasticity problems in each equations model. For the effect we carried out single equation tests.

ErPG : AR 1 – 4 F(4, 37) = 2.5266 [0.0570] CPIP : AR 1 – 4 F(4, 37) = 1.9897 [0.1163] iP : AR 1 – 4 F(4, 37) = 5.5068 [0.0014]** GDPP : AR 1 – 4 F(4, 37) = 7.6611 [0.0001]** ErPG : Normality Chi^2 (2) = 4.4699 [0.1070] CPIP : Normality Chi^2 (2) = 1.3496 [0.5093] iP : Normality Chi^2 (2) = 1.579 [0.4541] GDPP: Normality Chi^2 (2) = 0.56602 [0.7535] ErPG : ARCH 4 F(4, 33) = 0.28257 [0.8872] CPIP : ARCH 4 F(4, 33) = 0.15816 [0.9579] iP : ARCH 4 F(4, 33) = 0.2554 [0.9043] GDPP: ARCH 4 F(4, 33) = 0.21532 [0.9280] Vector AR 1 – 4 F(64,166) = 1.2095 [0.1697] Vector normality Chi^2 (8) = 8.3916 [0.3962]

6

From the obtained results it is possible to understand that, despite of the existing auto-correlation problems in the interest rate equations and of Portugal’s GDP, there aren’t any problems concerning normality, or in the ARCH tests, reaching the conclusion that the estimated model is a reasonable model.

Following the previous analysis, it was still done the graphic representation of the effective and estimated values of the variables throughout the time (left side of Figure 4) and the residuals scale of the model (right side of Figure 4).

Figure 4: Effective and Estimated Values and Residual Scale

4.8 ErPA Fitted rE rP A (scaled ) 2 4.6 0 4.4 1980 1985 1990 1995 2000 1980 1985 1990 1995 2000

IP CP Fitted rIP C P (scaled ) 5 2 4 0 3 1980 1985 1990 1995 2000 1980 1985 1990 1995 2000

0.3 iP Fitted 2 riP (scaled ) 0.2 0 0.1 -2 1980 1985 1990 1995 2000 1980 1985 1990 1995 2000

lnP IBP Fitted rln P IB P (scaled ) 4.75 2 4.50 0 4.25 1980 1985 1990 1995 2000 1980 1985 1990 1995 2000

As we can observe, the estimated model follows the evolution of the variables throughout several years.

On the other hand, it is not possible to find a pattern of behaviour in the residuals that suggests the existence of any form of auto-correlation, confirming that we are in presence of a reasonable model.

Afterwards, the stability of the model parameters was evaluated by using two Chow tests type normally known as 1-step Chow-tests and Break-point Chow-tests.

In this context, a recursive estimation was performed from an initial model with thirty six (36) observations. The stability tests were developed to a significance level of 1%.

In Figures 5 and 6 it is shown the output of the recursive analysis.

7 Figure 5: Recursive Estimation Statistics (1 Step Chow Tests)

1.0 1.0 1up ErPA 1% 1up IPCP 1%

0.5 0.5

0.0 0.0 1990 1995 2000 1990 1995 2000 1.0 1.0 1up iP 1% 1up lnPIBP 1%

0.5 0.5

0.0 1990 1995 2000 1990 1995 2000 1.0 1up CHOW s 1%

0.5

1990 1995 2000

Figure 6: Recursive Estimation Statistics (Break-Point Chow Tests)

1.0 1.0 Ndn ErPA 1% Ndn IPCP 1%

0.5 0.5

1990 1995 2000 1990 1995 2000 1.0 1.0 Ndn iP 1% Ndn lnPIBP 1%

0.5 0.5

1990 1995 2000 1990 1995 2000 1.0 Ndn CHOW s 1%

0.5

1990 1995 2000

As we can observe, for any of the four equations and for both tests, parameter stability is clearly acceptable.

It is just detected through the first test some instability concerning the real exchange rate, coinciding that situation with the collapse of the EMS in August 1993.

However, not even during this period, can we state that there was an exception to the parameters stability, since that the straight line that defines the significance limit of 1% was ever reached.

8 Finally, a dynamic analysis was performed to the roots of the model to evaluate its stability. The main results of this analysis are represented in Figure 7, corresponding to the Roots of Companion Matrix:

Figure 7: Stability of the Model (Roots of Companion Matrix)

1.00 Roots of companion matrix

0.75

0.50

0.25

0.00

-0.25

-0.50

-0.75

-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Figure 7 shows the non-existence of roots outside the companion matrix, confirming, thus, the stability of the model.

In this context, it was possible to go forward with a simulation analysis, from which we evaluated the main effects of the Portuguese disinflation process.

VI. Simulation Analysis

With the Simulation Analysis we simulated with the constructed model the existence of a unit variation in one of the endogenous variables, evaluating later on its impact (response) in the other endogenous variables.

This study intends, in a certain way, to clarify the influence that the real exchange rate appreciation may have over output and price levels, subjacent to the disinflation strategy.

We document in Figure 8 the results of a unit shock in each of the endogenous variables throughout eighty (80) periods. As the variables are in levels, the shock results always tend to be zero.

9

Figure 8: Simulation Analysis (80 periods)

1.0 0.20 0.04 ErPA (ErPA eqn) IPCP (ErPA eqn) iP (ErPA eqn) lnPIBP (ErPA eqn) 0.15 0.01 0.5 0.02 0.10 0.00 0.0 0.05 0.00 0.00 -0.01 0 50 0 50 0 50 0 50 1.00 0.5 ErPA (IPCP eqn) IPCP (IPCP eqn) 0.2 iP (IPCP eqn) 0.10 lnPIBP (IPCP eqn) 0.75 0.05 0.0 0.50 0.1 0.00 -0.5 0.25 0.0 0 50 0 50 0 50 0 50 1.0 ErPA (iP eqn) IPCP (iP eqn) 1.0 iP (iP eqn) lnPIBP (iP eqn) 0.5 2 0.5 0.0 1 0.5 0.0 0.0 -0.5 0 -0.5 0 50 0 50 0 50 0 50 1.0 0.5 ErPA (lnPIBP eqn) 0.75 IPCP (lnPIBP eqn) 0.50 iP (lnPIBP eqn) lnPIBP (lnPIBP eqn) 0.5 0.0 0.50 0.25 0.0 -0.5 0.25 0.00 0.00 -0.25 -0.5 0 50 0 50 0 50 0 50

For example, the first line of each graph shows the result of a unit variation in the real exchange rate over the other endogenous variables.

Positive variations in the real exchange rate of the Escudo to the Deutsch Mark are equivalent to the existence of real depreciations, whereas the negative variations correspond to real exchange rate appreciation. It must be noticed that the situation manifested in Figure 8 is exactly the opposite. However, we will opt by doing a symmetric analysis.

Through this analysis, we may conclude the following:

Firstly, it is verified that apart from the origin of the unit shock, price levels register a relatively strong general reduction tendency. Moreover, the most significant price levels reductions coincide with the periods of real exchange rate appreciation, output fall and drop of the interest rate.

Secondly, it is observed that the interest rate reaction tends to follow price levels behaviour when the shock is originated in the real exchange rate and in price levels. However, if the impulse has its origin in output, the interest rate reveals an anti-cycle behaviour in face of this variable. In the former case, the interest rate reaction is relatively weak, but the same does not happen when the shock is originated through output, registering, in this case, relatively strong variations in the interest rate.

10 Thirdly, it is verified that the real exchange rate tends to follow the output behaviour when the unit shock is originated on price levels and interest rate, that is, the real exchange rate appreciation phases coincide with output fall periods, while during the real exchange rate depreciation phases it is observed a growth in the level of economic activity. On the contrary, if the impulse is originated through output, the real exchange rate works in an anti-cycle way in face of this variable.

Finally, it is verified a general tendency for output fall in the cases where the unit shock takes place in the exchange rate and in price levels, although not significant. The output reaction, on the other hand, presents itself relatively strong if the impulse takes place in the interest rate, emphasizing, the sound functioning of the financial markets. In the latter case, it is possible to observe some relatively durable phases of output growth, reflecting the descending course of the capital cost.

VII. Conclusion

It was possible to conclude that the acceptance of the German monetary policy and the pegging of the Portuguese Escudo to the Deutsch mark allowed the Portuguese economy to experience a sustained disinflation process, as well as reaching their main aim, price stability.

However, associated to the gains reached in credibility and stability, the adoption of a disinflation policy led to a real appreciation of the Portuguese Escudo to the Deutsch Mark, limiting in this way output growth.

In face of this evidence, the aim was then to clarify the influence that the real exchange rate appreciation may have had over the output, the interest rate and on price levels.

Due to the continuous and gradual nature of the disinflation process, everything seems to indicate that the output costs may present themselves less significant and distributed throughout the period.

Moreover, Portugal may benefit from an increasing mobility of capitals and from a more efficient functioning of financial markets, allowing, indeed, a significant reduction in the interest rates, a phenomenon that may have worked in an anti-cycle way in face of the reduction of the level of economic activity.

The nominal stability policy of the Escudo may, in this manner, serve as an example for other small open economies that may be presently involved in the European integration process. In this context, in future studies it will be analysed the disinflation benefits, as well as the output costs of such policies in countries such as the Czech Republic, Estonia, Hungary, Poland and Slovenia. 11