I. Introduction

This project is motivated by the experience of the countries that devalued during the 1992-93 Exchange Rate Mechanism (ERM) crisis. Developments in some of these economies seem to refute the conventional wisdom about the inflationary effects of devaluations. Despite devaluations of the Spanish peseta and Portuguese escudo, the annual inflation rates in these countries continued to decline in 1993 and 1994. ^2/

This paper focuses on the Spanish case. The main question addressed below is whether all inflationary effects of the devaluations were offset by changes in exogenous factors, such as fiscal and monetary policies or foreign demand, which caused a recession in Spain. An additional issue that is raised is to what extent is the transmission of recent exchange rate changes into prices consistent with Spain’s long-run experience.

The term “pass-through” is predominantly used in the literature to denote effects of exchange rate changes on prices of particular goods. In this paper, however, it is used in a broader context, i.e. it applies to the effect of a devaluation on aggregate price indices.

Spain has been a member of the EC since 1986 and has participated in the fixed exchange rate system since 1989. The Spanish peseta was devalued three times in 1992-93 since the beginning of the ERM crisis by a cumulative amount of 21.4 percent, ^3/ which translated into a 14 percent annual depreciation in the nominal effective exchange rate by the end of 1993 (Chart 1). In spite of this, consumer price inflation was only 4.5 percent in 1993, which was lower than in 1992, and 4.5 percent again in 1994. ^4/ Assuming that there exists a one-to-one long-run relationship between exchange rates and prices, these developments are evidence of a more complex determination of inflation in the short-run.

Nominal Effective Exchange Rate

(1990=100)

Citation: IMF Working Papers 1996, 114; 10.5089/9781451943412.001.A001

Source: IMF, International Financial Statistics.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Nominal Effective Exchange Rate

(1990=100)

Citation: IMF Working Papers 1996, 114; 10.5089/9781451943412.001.A001

Source: IMF, International Financial Statistics.

• Download Figure
• Download figure as PowerPoint slide

Nominal Effective Exchange Rate

(1990=100)

Citation: IMF Working Papers 1996, 114; 10.5089/9781451943412.001.A001

Source: IMF, International Financial Statistics.

• Download Figure
• Download figure as PowerPoint slide

There is a widespread perception that devaluations have inflationary effects. This has been one reason for the tendency of governments to hesitate to use devaluations as a policy tool to correct external imbalances. ^5/ However, developments in Spain would seem to contradict this experience.

The issue of the inflationary impact of devaluations has been addressed previously. Theoretical contributions include, among others, Wilson (1976), Guitiàn (1976) and Rodriguez (1987). The empirical evidence on the degree of the exchange rates pass-through varies. Spitäller (1980) and Kreinin (1977) reported full pass-through in all but large economies. On the other hand, Cooper (1971) and Himarios (1987) found pass-through to be incomplete.

These earlier empirical studies explored only the short-run dynamic link between devaluation and inflation. In addition, they made assumptions about exogeneity and endogeneity of different variables without performing corresponding statistical tests. These problems have later been circumvented in papers which estimate both long-run as well as short-run relationships, by using cointegration analysis and error-correction models. However, the more recent literature tends to focus only on exchange rate pass-through into export prices (Sarker, 1993; Menon, 1992; Hung, Kim, Ohno, 1993; Athukorala and Menon, 1994) or import prices (Menon, 1993; Kim, 1990). De Grauwe and Tullio (1994) analyze the relationship of exchange rate changes and inflation rates across ERM countries that devalued in 1992-93. Although they do not incorporate wage-price dynamics into their model, they find supporting evidence that the recent low inflation experience may be attributed to the deflationary effect of high real interest rates.

Unlike the studies above, this paper focuses on the exchange rate pass-through not only into import and export prices, but into the aggregate price level as well. Since Spain’s historical unemployment rate was above the European average, it is necessary to determine whether recent changes in labor market conditions affected the disinflation process. Hence the behavior of wages and their impact on the price level are also included in the model.

The supply-side model in the main body of this paper is developed in accordance with the view that any information contained in determinants of aggregate demand, is accounted for by variables representing the cyclical position of the economy, such as capacity utilization, unemployment rate and productivity.

Long-run equilibrium relationships for import and export prices, nominal wages and the consumer price index are estimated using the Johansen maximum-likelihood cointegration method, which avoids the assumption of exogeneity of right-hand-side variables. The short-run dynamic behavior of the price indices is estimated with error-correction models.

Estimations suggest that Spain is a price-taker for imported goods. There is also support for the hypothesis of nominal wage and price rigidity.

The four-equation system is then simulated by calibrating shocks to data over 1992-93. This implies that any exogenous changes in the various determinants of aggregate demand and supply conditions, as well as any reverse causality from inflation to cyclical factors, are implicitly taken into account by making use of the actual values that resulted as equilibrium outcomes in that particular historical instance. The results indicate that the cost-push up effect of devaluations may have been completely offset by the determinants of the cyclical position of the economy and resulted in constant inflation over the subsequent two-year period.

The estimated model predicts well all price indices during the sample period. This implies that the relatively low inflation rate after substantial devaluations is not an unusual phenomenon in Spain and is in line with the historical experience.

The outline of the paper is as follows. Section 2 summarizes theories of inflation determination and discusses specific channels of the exchange rate pass-through. Section 3 presents the analytical framework. Section 4 briefly reviews the econometric methods used in the analysis. The empirical results are reported in section 5. Simulations of the contemporaneous impact of devaluations and shocks to cyclical variables are presented in Section 6. Section 7 summarizes the conclusions.

II. Theory and Background

As is well known, the origins of the exchange rate pass-through theory are embedded in the law of one price. Owing to spacial arbitrage, the price of a given good is the same (measured in a common currency) regardless of location. Purchasing power parity (PPP) theory assumes that the rule applies to all traded goods as long as the assumption of perfect competition in the domestic and foreign markets holds continuously. The pass-through theory typically imposes that the importing country be small relative to world markets—a price taker. In this case, the theory predicts full transmission of a devaluation into domestic prices of traded goods in the long run. Drawing on the assumptions listed above, both monetary (for example Dornbusch, 1973) and wage-setting models (Bruno, 1978) predict that domestic prices increase by the full amount of a devaluation. Empirical evidence on PPP is mixed: although in the long-run it appears to hold reasonably well, deviations are observed in the short run (Levich, 1985).

The direct impact of an exchange rate devaluation on inflation occurs through prices of imported goods and services. It is possible, however, that a change of the exchange rate is not passed on to consumer prices immediately, owing to the foreign exporters’ effort to maintain their market share. This phenomenon is known as “pricing to market behavior” (Krugman, 1987). It is an outcome of imperfect competition, lack of spacial arbitrage or imperfect substitutability of traded goods (Dornbusch, 1987). Pricing to market occurs when foreign firms allow import prices to rise less than the exchange rate change when that change is either unanticipated or expected to reverse.

Once prices of traded goods, the domestic price effect is determined by several factors (Goldstein and Khan, 1985). The first one is the substitutability of imports and domestically produced goods. If they were perfect substitutes, the home price level should be perfectly correlated with import prices. However, since the law of one price has been rejected in most empirical literature and the assumption of perfect substitutability does not hold, the expected correlation of prices should be less than unity (Kravis and Lipsey, 1978).

The second important factor is the degree of openness of the economy. According to production theory, the price elasticity with respect to import prices can be approximated by the value share of imports in total output, as in the case of the Cobb-Douglas production function. Therefore a more import-open economy implies a higher import-price effect. This reasoning assumes that the imports are intermediate goods used in domestic production rather than final goods.

The third element is the elasticity of nominal wages and other factor prices with respect to domestic price changes. If wages are set in a bargaining process, unions are likely to note increases in the overall cost of living and demand a rise in wages following a change in the exchange rate. Sachs (1980) claims that if wage indexation is widespread and the downward real wage resistance is strong, the domestic price effects of import price changes will likewise be strong, since firms would pass the increase in production costs on to the prices of final goods. Therefore, the higher the elasticity of domestic prices to changes in factor prices, the greater the elasticity of domestic prices with respect to import prices. Thus, in theory, a wage-price inflation spiral may ultimately undercut gains from devaluation.

The direct increase in relative prices of tradables tends to induce a change in consumption patterns from traded toward nontraded goods and to add to price inflation of the latter. This is the expenditure-switching effect of devaluation (Edwards, 1988). In addition, if the country is a net borrower, devaluation reduces the real value of assets denominated in terms of domestic currency. This negative wealth effect causes expenditure reduction. Prices of imported final goods are not the only ones affected by a devaluation. An increase in the cost of imported intermediate goods can push up prices on non-traded goods as well and contribute to inflationary pressures. On the export side, exchange rate pass-through is the degree to which exchange rate changes affect the prices of domestic exports in foreign currency terms. Home export prices tend to rise to offset at least some of the reduction of the foreign-currency price of exports (Goldstein and Khan, 1985). Devaluations lead to higher export prices as home goods become more competitive in foreign markets and as exporters increase the markup for their products. ^6/

Concerning the other determinants of inflation, the empirical section of this paper does not include monetary variables, and relies instead on cyclical indicators such as capacity utilization and the unemployment rate, and some determinants of the aggregate supply approach, such as the determinants of prices as markups over labor costs, adjusted for productivity and other production costs, such as imported inputs. ^7/

III. Analytical Framework

The model of price determination presented in this section includes both foreign and domestic price components. Import and export prices are estimated to capture direct channels of exchange rate pass-through. To assess the impact of devaluation on the general price level, nominal wage and consumer price index equations are added. ^8/

The import price (P_M) is by construction an index of import prices of goods and services. It is an increasing function of the effective nominal exchange rate (E) and the foreign currency price of exports to Spain (P_X*). The foreign export price is an average of foreign partners export deflators, weighted by the export share, and is measured in foreign currency. Another explanatory variable, the price of oil (P_OIL), is introduced to account for supply side-shocks because of the large oil content in Spanish imports. An increase in the price of oil pushes up domestic prices of imported goods. The cyclical behavior of the economy is represented by capacity utilization (CU), defined as a percentage of total capacity being used. The error term of the equation is labeled ε_M. The long-run import price relationship to be estimated is specified in equation (1), with the expected long-run signs of coefficients denoted below. It is important to note that a coefficient α_i can be interpreted as the elasticity of the import price with respect to the i-th exogenous variable.

The export price (P_X) is measured as an index of export prices of goods and services (equation 2). Following the specification adopted by Menon (1992), Hung et. al. (1993) and Athukorala and Menon (1994), the export price includes a variable markup over the unit costs of production and results from profit-maximizing behavior by the producers. This effect is captured by inclusion of unit labor costs (ULC), which are defined as the total compensation of employees per unit of gross domestic product at market prices, or as the compensation per worker times the inverse of productivity. ^9/ The markup depends on demand pressures in domestic markets and on competitive pressures in the world market. The world price in home currency is approximated by the effective exchange rate (E) multiplied by the price of imports of Spain’s trading partners (P_M*). The latter is a weighted average of foreign partners’ import deflators measured in foreign currency, weighted by import shares. If exporters are price-takers, then β₁ will equal unity and there will be no pricing to market. However, if exporters faced no competition on the world market, they would not adjust prices in home currency and exchange rate changes would be fully reflected in foreign currency prices (β₁=0). According to Deppler and Ripley (1978), small open economies tend to base their export prices on competitors’ prices rather than on domestic factor costs. This implies an expected relationship among coefficients: β₃ < β₁ + β₂. ε_X is the error term. The cointegrating equation and the expected long-run signs of coefficients are:

Firms and unions bargain over the real wage (W-P), ^10/ which is expected to rise with falling unemployment rate (UR). Net wages, excluding taxes and social security contributions, are used in the analysis. ^11/ The real wage depends positively on productivity (PR). Productivity is measured as value added per worker (real GDP divided by total employment). Thus the long-run relationship between real wages and their determinants are given by equation (3), followed by the expected signs of coefficients. Again, the error term is ε_w.

The domestic price level (P), represented by the consumer price index, is by definition a weighted index of prices of imported (P_M) and non-traded goods. Following the specification of Mellis (1993) and Davis and Jensen (1994) prices are determined by wages and import prices. Prices of home goods are determined by the cost mark up over unit labor costs (ULC), which include both wage and non-wage shares of employers’ outlays. Capacity utilization (CU) is included in order to correct for aggregate demand conditions, which can indicate variability in profit margins. In other words, the markup of prices over labor and import costs is expected to rise with the level of economic activity. The error term in equation (4) is ε_p.

Although monetary policy and interest rates are not incorporated in the wage and price equation directly, it is possible to identify the channels of their impact in this model. Monetary policy affects the exchange rate and therefore prices of traded goods.

IV. Econometric Techniques and Data

The long-run equilibrium (cointegrating) relationships among variables may be obtained if individual variables are integrated of same order, since any divergence of cointegrated variables must be stochastic and must diminish over time. Therefore presence of non-stationarity is tested for each variable, using the Dickey-Fuller (DF) and Augmented Dickey-Fuller (ADF) tests.

The cointegrating relationships are estimated using the Johansen’s maximum-likelihood method. ^12/ The short-run dynamics are estimated using an error-correction model. Such models capture the adjustment of the dependent variable to the deviation of the explanatory variables from the equilibrium long-run relationship. To determine the dynamic specification, a Hendry general-to-specific approach is applied. The residuals from the cointegrating equation are used together with lagged independent variables as regressors. Insignificant values of dynamic variables are progressively eliminated. A constant term is also added to these dynamic equations. In the case when its presence is justified, it implies a deterministic trend in the deterministic equation under question.

The analysis is performed on quarterly data from the OECD Analytical Database. The sample used in the analysis ranges from the first quarter of 1960 to third quarter of 1993. All variables were seasonally unadjusted and transformed into logarithms, unless specified otherwise. Lower letters stand for variables in logarithm form. Due to the log-linear specification of the cointegrating equations, the estimated coefficients can be interpreted as long-run elasticities.

In addition, several dummy variables were tested in dynamic equations. Included dummy variables represent: seasonal variations (DUM1, DUM2, DUM3, DUM4); entry into EC in 1986 and elimination of import tariffs (DUMEC); implementation of the Moncloa Agreement in 1977 (DUMM); entry of Spain into the ERM system in 1989 (DUMERM); fixed and flexible exchange rate systems (DUMFIX).

V. Empirical Results

The presence of nonstationarity of the individual series was tested using the Dickey-Fuller (DF) and Augmented Dickey-Fuller (ADF) tests. The t-statistics of the two tests are reported in Table 1. ^13/ Critical values for tests were computed using the response surface estimates given by MacKinnon (1990). The last column gives the implied order of integration. Nominal wages, prices and unemployment appeared to be integrated of order two, ^14/ while the other variables were integrated of order one. Capacity utilization should be stationary by definition. However, due to small-sample properties of the ADF test, the series seem to be non-stationary in level and are treated as integrated of order one.

Table 1.

Tests for Stationarity

Notes: Lower letters—logarithm; d—difference. Critical values at 95 percent confidence level for a/ DF with constant and trend at 95 percent confidence level is -3.4472; b/ ADF with constant, trend and four lags is 3.4450.

Table 1.

Tests for Stationarity

		Level		First difference
Variable		DF	ADF	DF	ADF	~I()
p		-1.398	-1.612	-5.726	-2.474	I(2)
	dp			-19.65	-7.630
w		3.103	-0.830	-3.854	-1.750	I(2)
	dw			-11.89	-4.891
pM		-0.689	-1.368	-5.134	-3.637	I(1)
pX		-1.512	-1.273	-13.49	-3.511	I(1)
UR		-1.377	-3.425	-1.590	-2.181	I(2)
	dUR			-5.355	-4.413
CU		-3.267	-3.721	-14.25	-4.805	I(1)
pr		-3.083	-2.720	-12.53	-4.912	I(1)
pOIL		-1.102	-0.986	-11.23	-5.602	I(1)
ulc		-2.817	-2.951	-10.66	-4.664	I(1)
e		-1.834	-2.236	-8.971	-4.838	I(1)
pX*		-0.116	-1.904	-9.992	-4.039	I(1)
pM*		-0.415	-1.478	-10.20	-4.312	I(1)
w-p		-0.866	-2.514	-8.075	-4.715	I(1)
e+PM*		-1.154	-1.583	-9.801	-4.251	I(1)
e+pX*		-1.376	-1.550	-8.950	-3.982	I(1)

Notes: Lower letters—logarithm; d—difference. Critical values at 95 percent confidence level for a/ DF with constant and trend at 95 percent confidence level is -3.4472; b/ ADF with constant, trend and four lags is 3.4450.

View Table

Table 1.

Tests for Stationarity

		Level		First difference
Variable		DF	ADF	DF	ADF	~I()
p		-1.398	-1.612	-5.726	-2.474	I(2)
	dp			-19.65	-7.630
w		3.103	-0.830	-3.854	-1.750	I(2)
	dw			-11.89	-4.891
pM		-0.689	-1.368	-5.134	-3.637	I(1)
pX		-1.512	-1.273	-13.49	-3.511	I(1)
UR		-1.377	-3.425	-1.590	-2.181	I(2)
	dUR			-5.355	-4.413
CU		-3.267	-3.721	-14.25	-4.805	I(1)
pr		-3.083	-2.720	-12.53	-4.912	I(1)
pOIL		-1.102	-0.986	-11.23	-5.602	I(1)
ulc		-2.817	-2.951	-10.66	-4.664	I(1)
e		-1.834	-2.236	-8.971	-4.838	I(1)
pX*		-0.116	-1.904	-9.992	-4.039	I(1)
pM*		-0.415	-1.478	-10.20	-4.312	I(1)
w-p		-0.866	-2.514	-8.075	-4.715	I(1)
e+PM*		-1.154	-1.583	-9.801	-4.251	I(1)
e+pX*		-1.376	-1.550	-8.950	-3.982	I(1)

Notes: Lower letters—logarithm; d—difference. Critical values at 95 percent confidence level for a/ DF with constant and trend at 95 percent confidence level is -3.4472; b/ ADF with constant, trend and four lags is 3.4450.

Based on the unit-root tests for all the variables, the existence of long-run cointegrating equilibria is tested in the next step. The Johansen maximum-likelihood method was employed for this purpose, in particular the trended case, with no trend in the data-generating process. Table 2 indicates the number of cointegrating vectors for variables P_M, P_X, W, and P. Cointegration likelihood-ratio tests based on the trace of the stochastic matrix are reported. Two cointegrating vectors were found for each aggregate price index. Since the results suggest existence of more than one vector a single vector was selected on the basis of economic meaningfulness and bearing in mind considerations of Johansen and Juselius (1990).

Table 2.

Johansen Maximum Likelihood Tests for Cointegration

Notes: The number of cointegrating vectors is determined sequentially. r = number of cointegrating vectors, m = number of variables in VAR.1. If r=0 hypothesis is rejected, test the hypothesis that r<=1.2. If r=m, hypothesis that the variables in the VAR have a stationary process cannot be rejected.3. The test results provide evidence in favor of cointegration only in the case where 0<r<m.

Table 2.

Johansen Maximum Likelihood Tests for Cointegration

Cointegration LR Test Based on Trace of the Stochastic Matrix
Import prices List of eigenvalues:	.21640	.19040	.072854	.022170
Null	Alternative	Statistic	95%	90%
			Critical value
r=0	r>=1	58.6333	48.2800	45.2290
r<=1	r>=2	32.7839	31.5250	28.7090
r<=2	r>=3	10.3947	17.9530	15.6630
Export prices List of eigenvalues:	.16822	.094559	.043323
Null	Alternative	Statistic	95%	90%
			Critical value
r=0	r>=1	41.3033	31.5250	28.7090
r<=1	r>=2	18.0965	17.9530	15.6630
r<=2	r>=3	5.5804	8.1760	6.5030
Wages List of eigenvalues:	.28349	.20659	.012088
Null	Alternative	Statistic	95%	90%
			Critical value
r=0	r>=1	48.4632	31.5250	28.7090
r<=1	r>=2	20.4609	17.9530	15.6630
r<=2	r>=3	1.0216	8.1760	6.5030
Prices List of eigenvalues:	.43566	.19642	.11175	.014022
Null	Alternative	Statistic	95%	90%
			Critical value
r=0	r>=1	86.8002	48.2800	45.2290
r<=1	r>=2	33.0228	31.5250	28.7090
r<=2	r>=3	12.4668	17.9530	15.6630

Notes: The number of cointegrating vectors is determined sequentially. r = number of cointegrating vectors, m = number of variables in VAR.1. If r=0 hypothesis is rejected, test the hypothesis that r<=1.2. If r=m, hypothesis that the variables in the VAR have a stationary process cannot be rejected.3. The test results provide evidence in favor of cointegration only in the case where 0<r<m.

View Table

Table 2.

Johansen Maximum Likelihood Tests for Cointegration

Cointegration LR Test Based on Trace of the Stochastic Matrix
Import prices List of eigenvalues:	.21640	.19040	.072854	.022170
Null	Alternative	Statistic	95%	90%
			Critical value
r=0	r>=1	58.6333	48.2800	45.2290
r<=1	r>=2	32.7839	31.5250	28.7090
r<=2	r>=3	10.3947	17.9530	15.6630
Export prices List of eigenvalues:	.16822	.094559	.043323
Null	Alternative	Statistic	95%	90%
			Critical value
r=0	r>=1	41.3033	31.5250	28.7090
r<=1	r>=2	18.0965	17.9530	15.6630
r<=2	r>=3	5.5804	8.1760	6.5030
Wages List of eigenvalues:	.28349	.20659	.012088
Null	Alternative	Statistic	95%	90%
			Critical value
r=0	r>=1	48.4632	31.5250	28.7090
r<=1	r>=2	20.4609	17.9530	15.6630
r<=2	r>=3	1.0216	8.1760	6.5030
Prices List of eigenvalues:	.43566	.19642	.11175	.014022
Null	Alternative	Statistic	95%	90%
			Critical value
r=0	r>=1	86.8002	48.2800	45.2290
r<=1	r>=2	33.0228	31.5250	28.7090
r<=2	r>=3	12.4668	17.9530	15.6630

Notes: The number of cointegrating vectors is determined sequentially. r = number of cointegrating vectors, m = number of variables in VAR.1. If r=0 hypothesis is rejected, test the hypothesis that r<=1.2. If r=m, hypothesis that the variables in the VAR have a stationary process cannot be rejected.3. The test results provide evidence in favor of cointegration only in the case where 0<r<m.

Estimates of cointegrating vectors together with the dynamic equations are reported in Table 3.

Table 3.

Estimates of Cointegrating and Dynamic Equations

Notes:Only variables with coefficients significant at 95 confidence level are included in the dynamic equations presented in this table.t-statistics are reported in parenthesesd — denotes first difference of a variable

Table 3.

Estimates of Cointegrating and Dynamic Equations

${\left(5\right)\mathrm{p}}_{\mathrm{M}}\mathrm{=}\mathrm{.96}\mathrm{(}\mathrm{e}\mathrm{+}{{\mathrm{p}}_{\mathrm{X}}}^{*}\mathrm{)}\mathrm{+}\mathrm{.21}{\mathrm{p}}_{\text{O}\mathrm{IL}}\mathrm{+}\mathrm{.04}\mathrm{CU}\mathrm{\left(}\text{-}\mathrm{4}\mathrm{\right)}$
$\begin{array}{l}\left(6\right){\mathrm{dp}}_{\mathrm{M}}\mathrm{=}\underset{\mathrm{\left(}\mathrm{2.0}\mathrm{\right)}}{\mathrm{.02}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{8.0}\mathrm{\right)}}{\mathrm{.74}{\mathrm{dp}}_{\mathrm{M}}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{3.2}\mathrm{\right)}}{\mathrm{.25}{\mathrm{dp}}_{\mathrm{M}}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.5}\mathrm{\right)}}{\mathrm{.18}\mathrm{de}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.3}\mathrm{\right)}}{\mathrm{.13}\mathrm{de}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{4.0}\mathrm{\right)}}{\mathrm{.54}{{\mathrm{dp}}_{\text{X}}}^{*}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.2}\mathrm{\right)}}{\mathrm{.03}{\mathrm{dp}}_{\mathrm{OIL}}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.0}\mathrm{\right)}}{\mathrm{.002}\mathrm{dCU}\left(\text{-2}\right)}\text{\hspace{0.5em}-\hspace{0.5em}\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{2.0}\mathrm{\right)}}{\mathrm{.03}{\mathrm{ϵ}}_{\mathrm{M}}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\end{array}$
R-Squared = 0.81	DW = 2.1	F(8,96) = 50.31
$\begin{array}{l}\left(7\right){{\mathrm{p}}_{\mathrm{X}}}_{\mathrm{=}\mathrm{.55}\text{(}\mathrm{e}\mathrm{+}{{\mathrm{p}}_{\mathrm{M}}}^{*}\text{)\hspace{0.5em}}\mathrm{+}\mathrm{.35}\mathrm{ulc}}\end{array}$
${\left(8\right)\mathrm{dp}}_{\mathrm{X}}\mathrm{=}\underset{\mathrm{\left(}\mathrm{2.5}\mathrm{\right)}}{\mathrm{.18}}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{5.7}\mathrm{\right)}}{\mathrm{.47}{\mathrm{dp}}_{\mathrm{X}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.8}\mathrm{\right)}}{\mathrm{.31}\mathrm{de}}\text{\hspace{0.5em}+\hspace{0.5em}}\underset{\mathrm{\left(}\mathrm{3.1}\mathrm{\right)}}{\mathrm{.26}{{\mathrm{dp}}_{\mathrm{M}}}^{*}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.8}\mathrm{\right)}}{\mathrm{.24}{{\mathrm{dp}}_{\mathrm{M}}}^{*}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.7}\mathrm{\right)}}{\mathrm{.23}{{\mathrm{dp}}_{\mathrm{M}}}^{*}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.6}\mathrm{\right)}}{\mathrm{.07}\mathrm{dulc}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.1}\mathrm{\right)}}{\mathrm{.08}\mathrm{dulc}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\text{\hspace{0.5em}-}\underset{\mathrm{\left(}\mathrm{2.2}\mathrm{\right)}}{\mathrm{.04}{\mathrm{ϵ}}_{\mathrm{X}}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}$
R-Squared= 0.38	DW = 2.02	F(8,115) = 8.95
$\begin{array}{l}\left(9\right)\mathrm{\left(}\mathrm{w}\text{\hspace{0.5em}-\hspace{0.5em}}\mathrm{p}\mathrm{\right)}\mathrm{=}\mathrm{1.09}\mathrm{pr}\text{\hspace{0.5em}-\hspace{0.5em}}\mathrm{.01}\mathrm{UR}\end{array}$
$\begin{array}{l}\left(10\right)\mathrm{dw}\mathrm{=}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{(}\mathrm{-}\mathrm{3.1}\mathrm{)}}{\mathrm{.17}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{14.4}\mathrm{\right)}}{\mathrm{1.2}\mathrm{dw}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{4.2}\mathrm{\right)}}{\mathrm{.35}\mathrm{dw}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.1}\mathrm{\right)}}{\mathrm{.12}\mathrm{dp}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{15.2}\mathrm{\right)}}{\mathrm{.97}\mathrm{dpr}}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{10.8}\mathrm{\right)}}{\mathrm{1.2}\mathrm{dpr}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{4.2}\mathrm{\right)}}{\mathrm{.43}\mathrm{dpr}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{3.1}\mathrm{\right)}}{\mathrm{.04}{\mathrm{ϵ}}_{\mathrm{W}}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\text{-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{2.3}\mathrm{\right)}}{\mathrm{.004}\mathrm{DUMERM}}\end{array}$
R-Squared = 0.96	DW-statistic = 1.72	F(8,73) = 199.50
$\begin{array}{l}\left(11\right)\mathrm{p}\mathrm{=}\mathrm{.09}\mathrm{pm}\mathrm{+}\mathrm{.94}\mathrm{ulc}\mathrm{+}\mathrm{.006}\mathrm{CU}\end{array}$
$\begin{array}{l}\left(12\right)\mathrm{dp}\mathrm{=}\underset{\mathrm{\left(}\mathrm{3.0}\mathrm{\right)}}{\mathrm{.25}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.0}\mathrm{\right)}}{\mathrm{.21}{\mathrm{dp}}_{\mathrm{M}}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{2.8}\mathrm{\right)}}{\mathrm{.31}{\mathrm{dp}}_{\mathrm{M}}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.7}\mathrm{\right)}}{\mathrm{.26}{\mathrm{dp}}_{\mathrm{M}}}\mathrm{\left(}\text{-}\mathrm{3}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.7}\mathrm{\right)}}{\mathrm{.32}\mathrm{dulc}}\mathrm{\left(}\text{-}\mathrm{4}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.5}\mathrm{\right)}}{\mathrm{.001}}\mathrm{dCU}\text{-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{2.9}\mathrm{\right)}}{\mathrm{.07}\mathrm{ϵp}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\end{array}$
R-Squared = 0.65	DW = 1.71	F(6,87) = 27.3

Notes:Only variables with coefficients significant at 95 confidence level are included in the dynamic equations presented in this table.t-statistics are reported in parenthesesd — denotes first difference of a variable

View Table

Table 3.

Estimates of Cointegrating and Dynamic Equations

${\left(5\right)\mathrm{p}}_{\mathrm{M}}\mathrm{=}\mathrm{.96}\mathrm{(}\mathrm{e}\mathrm{+}{{\mathrm{p}}_{\mathrm{X}}}^{*}\mathrm{)}\mathrm{+}\mathrm{.21}{\mathrm{p}}_{\text{O}\mathrm{IL}}\mathrm{+}\mathrm{.04}\mathrm{CU}\mathrm{\left(}\text{-}\mathrm{4}\mathrm{\right)}$
$\begin{array}{l}\left(6\right){\mathrm{dp}}_{\mathrm{M}}\mathrm{=}\underset{\mathrm{\left(}\mathrm{2.0}\mathrm{\right)}}{\mathrm{.02}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{8.0}\mathrm{\right)}}{\mathrm{.74}{\mathrm{dp}}_{\mathrm{M}}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{3.2}\mathrm{\right)}}{\mathrm{.25}{\mathrm{dp}}_{\mathrm{M}}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.5}\mathrm{\right)}}{\mathrm{.18}\mathrm{de}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.3}\mathrm{\right)}}{\mathrm{.13}\mathrm{de}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{4.0}\mathrm{\right)}}{\mathrm{.54}{{\mathrm{dp}}_{\text{X}}}^{*}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.2}\mathrm{\right)}}{\mathrm{.03}{\mathrm{dp}}_{\mathrm{OIL}}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.0}\mathrm{\right)}}{\mathrm{.002}\mathrm{dCU}\left(\text{-2}\right)}\text{\hspace{0.5em}-\hspace{0.5em}\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{2.0}\mathrm{\right)}}{\mathrm{.03}{\mathrm{ϵ}}_{\mathrm{M}}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\end{array}$
R-Squared = 0.81	DW = 2.1	F(8,96) = 50.31
$\begin{array}{l}\left(7\right){{\mathrm{p}}_{\mathrm{X}}}_{\mathrm{=}\mathrm{.55}\text{(}\mathrm{e}\mathrm{+}{{\mathrm{p}}_{\mathrm{M}}}^{*}\text{)\hspace{0.5em}}\mathrm{+}\mathrm{.35}\mathrm{ulc}}\end{array}$
${\left(8\right)\mathrm{dp}}_{\mathrm{X}}\mathrm{=}\underset{\mathrm{\left(}\mathrm{2.5}\mathrm{\right)}}{\mathrm{.18}}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{5.7}\mathrm{\right)}}{\mathrm{.47}{\mathrm{dp}}_{\mathrm{X}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.8}\mathrm{\right)}}{\mathrm{.31}\mathrm{de}}\text{\hspace{0.5em}+\hspace{0.5em}}\underset{\mathrm{\left(}\mathrm{3.1}\mathrm{\right)}}{\mathrm{.26}{{\mathrm{dp}}_{\mathrm{M}}}^{*}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.8}\mathrm{\right)}}{\mathrm{.24}{{\mathrm{dp}}_{\mathrm{M}}}^{*}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.7}\mathrm{\right)}}{\mathrm{.23}{{\mathrm{dp}}_{\mathrm{M}}}^{*}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.6}\mathrm{\right)}}{\mathrm{.07}\mathrm{dulc}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.1}\mathrm{\right)}}{\mathrm{.08}\mathrm{dulc}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\text{\hspace{0.5em}-}\underset{\mathrm{\left(}\mathrm{2.2}\mathrm{\right)}}{\mathrm{.04}{\mathrm{ϵ}}_{\mathrm{X}}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}$
R-Squared= 0.38	DW = 2.02	F(8,115) = 8.95
$\begin{array}{l}\left(9\right)\mathrm{\left(}\mathrm{w}\text{\hspace{0.5em}-\hspace{0.5em}}\mathrm{p}\mathrm{\right)}\mathrm{=}\mathrm{1.09}\mathrm{pr}\text{\hspace{0.5em}-\hspace{0.5em}}\mathrm{.01}\mathrm{UR}\end{array}$
$\begin{array}{l}\left(10\right)\mathrm{dw}\mathrm{=}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{(}\mathrm{-}\mathrm{3.1}\mathrm{)}}{\mathrm{.17}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{14.4}\mathrm{\right)}}{\mathrm{1.2}\mathrm{dw}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{4.2}\mathrm{\right)}}{\mathrm{.35}\mathrm{dw}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.1}\mathrm{\right)}}{\mathrm{.12}\mathrm{dp}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{15.2}\mathrm{\right)}}{\mathrm{.97}\mathrm{dpr}}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{10.8}\mathrm{\right)}}{\mathrm{1.2}\mathrm{dpr}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{4.2}\mathrm{\right)}}{\mathrm{.43}\mathrm{dpr}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{3.1}\mathrm{\right)}}{\mathrm{.04}{\mathrm{ϵ}}_{\mathrm{W}}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\text{-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{2.3}\mathrm{\right)}}{\mathrm{.004}\mathrm{DUMERM}}\end{array}$
R-Squared = 0.96	DW-statistic = 1.72	F(8,73) = 199.50
$\begin{array}{l}\left(11\right)\mathrm{p}\mathrm{=}\mathrm{.09}\mathrm{pm}\mathrm{+}\mathrm{.94}\mathrm{ulc}\mathrm{+}\mathrm{.006}\mathrm{CU}\end{array}$
$\begin{array}{l}\left(12\right)\mathrm{dp}\mathrm{=}\underset{\mathrm{\left(}\mathrm{3.0}\mathrm{\right)}}{\mathrm{.25}}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.0}\mathrm{\right)}}{\mathrm{.21}{\mathrm{dp}}_{\mathrm{M}}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\text{\hspace{0.5em}-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{2.8}\mathrm{\right)}}{\mathrm{.31}{\mathrm{dp}}_{\mathrm{M}}}\mathrm{\left(}\text{-}\mathrm{2}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{3.7}\mathrm{\right)}}{\mathrm{.26}{\mathrm{dp}}_{\mathrm{M}}}\mathrm{\left(}\text{-}\mathrm{3}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.7}\mathrm{\right)}}{\mathrm{.32}\mathrm{dulc}}\mathrm{\left(}\text{-}\mathrm{4}\mathrm{\right)}\mathrm{+}\underset{\mathrm{\left(}\mathrm{2.5}\mathrm{\right)}}{\mathrm{.001}}\mathrm{dCU}\text{-\hspace{0.5em}}\underset{\mathrm{\left(}\text{-}\mathrm{2.9}\mathrm{\right)}}{\mathrm{.07}\mathrm{ϵp}}\mathrm{\left(}\text{-}\mathrm{1}\mathrm{\right)}\end{array}$
R-Squared = 0.65	DW = 1.71	F(6,87) = 27.3

Notes:Only variables with coefficients significant at 95 confidence level are included in the dynamic equations presented in this table.t-statistics are reported in parenthesesd — denotes first difference of a variable

The long-run elasticity of import prices with respect to the home currency price of foreign exports is close to unity, as expected (equation 5, Table 3). Coefficients α₁ and α₂ were imposed to be identical, as there is no strong a priori reason to expect that the effects of the exchange rate and foreign export prices onto import prices should differ in the long run. Also, higher prices of oil imply higher imports prices, as in a case of a supply shock. Larger lagged capacity utilization in the home economy raises import prices as well. This effect is, however, relatively small.

According to the dynamic equation (6), deviations from the long-run equilibrium are corrected only very slowly, about 3 percent occurring in the first quarter. ^15/ Also, 74 percent of previous period’s changes are transmitted into the current quarter. The immediate responsiveness of import price inflation to changes in the exchange rate is smaller than the responsiveness to foreign export price changes. This could provide support for the pricing to market hypothesis. Supply shocks, resulting in increased prices of oil have only a small effect on import prices. Changes in the cyclical position of the economy, represented by the capacity utilization, appear to have a marginal impact on inflation. All seasonal and structural dummy variables were insignificant.

To obtain the cointegrating vector for the export price, inclusion of a dummy variable representing accession to the EC in 1986 (DUMEC) is required. This shift in Spain’s export price determination could be attributed to the elimination of import tariffs by EC member countries. In the cointegrating regression (equation 7), identical coefficients were imposed a priori on the effective exchange rate and the foreign import prices. The elasticity of export prices with respect to unit labor costs is smaller than with respect to foreign import prices. ^16/

The coefficient of the error-correction term is negative in the dynamic equation (8), but of a small magnitude. It implies slow adjustment, since only four percent of the deviations from the long run equilibrium are reversed in the following quarter. Current changes in the exchange rate have a slightly larger effect on current inflation than changes in current foreign export prices. However, past exchange rate changes are insignificant, while about 25 percent of increases in prices of foreign exports from two past quarters still have an impact. This could, as in the case of import prices, provide evidence for pricing to market behavior. Lagged increases in unit labor costs also contribute to export price inflation. None of the seasonal or structural dummy variables were significant.

A complication arises in the estimation of the cointegrating vector of nominal wages, because the net nominal wages and prices are integrated of order two. Following the error-correction specification of Hall and Henry (1987), a real wage variable (w-p) is constructed and used in the cointegrating equation. This is equivalent to imposing a unitary long-run elasticity of nominal wages with respect to prices. However, change in nominal wage is used as the dependent variable in the dynamic equation, where the change in the price level is an independent variable.

The long-run relationship indicates that the elasticity of the real wages with respect to productivity is also unity (equation 9). This implies that employees benefit from productivity in such a way as to leave profit shares constant. On the other hand, the unemployment rate lowers the real wage only marginally. According to Dreze and Bean (1990) this can be interpreted as a potential for persistent unemployment, possibly as a result of the ineffectiveness of the unemployed outsiders (Layard and Bean, 1989). These results confirm conventional conclusions about the Phillips curve specification of wage determination amended to include the target-real-wage hypothesis.

Deviations from the long-run equilibrium are corrected very slowly—only 4 percent during the first quarter (equation 10). Therefore, there is indication of nominal wage rigidity in Spain. Only current price inflation passes through to wage inflation. This may imply that the current changes in prices are a reasonable proxy for all future and past price changes that are being considered in the bargaining process. Current and lagged increases in productivity contribute to wage growth. Lagged changes in the unemployment rate appear not to have a significant effect on wage inflation. ^17/ This could be an indication that labor market reform in Spain is necessary to introduce responsiveness of wages to labor market conditions. ^18/ Wage inflation has slowed slightly after the entry into ERM judging by the significance of the dummy variable (DUMERM). Neither seasonal, nor other structural dummies were significant.

The cointegrating vector for the consumer price index (equation 11) confirms the hypothesis that higher import prices and unit labor costs push up the price level. The effect of labor cost is clearly dominant. The long-run elasticity of prices with respect to import goods prices is very small. ^19/ The sum of the two coefficients is close to unity. The capacity utilization has a small positive effect on prices in the long run. This pro-cyclical type of markup indicates that prices increase only slightly beyond the level specified by wages and import prices in a boom.

Only 7 percent of the deviation from the long run equilibrium are corrected in the first quarter (equation 12). This indicates nominal price rigidity in the short-run. The response of prices to import prices in the short-run is quite notable. ^20/ Past changes in unit labor cost also increases the inflation rate. The results show that changes in capacity utilization contribute only marginally to price changes.

The results indicate long-run homogeneity in all cointegrating vectors, with the possible exception of export prices. There is a very slow adjustment towards the long-run equilibrium in all dynamic equations.

VI. Simulations

1. Joint simulation

This section attempts to answer the main question raised in the paper, that is, whether the shocks of the early 1990s to determinants of the cyclical position of the economy eliminated any additional inflationary impact resulting from devaluations.

First, for each period, actual values of the four exogenous variables (nominal effective exchange rate, unemployment rate, capacity utilization, and productivity) are imposed on the model estimated in section V, yielding baseline series for the simulated variables (CPI, wages, import and export prices). ^{21/ 22/}

Second, the model is simulated by adding to the exogenous variables shocks designed to represent the contemporaneous shocks that occurred in Spain from Q4 1992 till Q3 1993. For this reason, the magnitudes of these shocks correspond to averages of the actual values during this period. They are as follows: a 5.3 percent depreciation of the exchange rate; a 0.5 percentage point decrease in the capacity utilization and a 1.2 percentage point increase in the unemployment rate associated with the ongoing recession; and a 0.8 percent increase in the productivity caused by a sharper decline in employment that in GDP. ^23/

The simulation should be interpreted as representing how much more inflation Spain should expect compared to a baseline, following changes in the exchange rate and shocks to other determinants of the aggregate demand and supply. The results are presented in Table 4.

Table 4.

Model Response to Exogenous Shocks

(Percent deviation from the baseline value after 4 and 8 quarters)

Notes: The shocks include: a 5.3 percent devaluation of the nominal effective exchange rate; a 0.5 percentage point decrease in capacity utilization; a 1.2 percentage points increase in the unemployment rate; 0.8 percent increase in productivity.

Table 4.

Model Response to Exogenous Shocks

(Percent deviation from the baseline value after 4 and 8 quarters)

	Import price	Export price	Wages	Prices
4 Qtrs.	3.0	1.2	0.6	0.1
8 Qtrs.	3.1	1.3	-0.3	0.0

Notes: The shocks include: a 5.3 percent devaluation of the nominal effective exchange rate; a 0.5 percentage point decrease in capacity utilization; a 1.2 percentage points increase in the unemployment rate; 0.8 percent increase in productivity.

View Table

Table 4.

Model Response to Exogenous Shocks

(Percent deviation from the baseline value after 4 and 8 quarters)

	Import price	Export price	Wages	Prices
4 Qtrs.	3.0	1.2	0.6	0.1
8 Qtrs.	3.1	1.3	-0.3	0.0

Notes: The shocks include: a 5.3 percent devaluation of the nominal effective exchange rate; a 0.5 percentage point decrease in capacity utilization; a 1.2 percentage points increase in the unemployment rate; 0.8 percent increase in productivity.

The effects of the devaluation on import prices are partially mitigated by the decreasing capacity utilization. As a result, the import prices increase by about 3 percent in eight quarters despite the ongoing recession. Most of the exchange rate pass-through occurs within one year. Export prices react to exogenous shocks more gradually: the increase is only 1.3 percent in two years. As in the case of import prices, most of the pass-through takes place in the first year.

Previous studies, such as Spitäller (1980) or Kreinin (1977) have found shorter lags in the import price pass-through. However, both analyses confirm that there is a marked tendency for import prices to rise more quickly than export prices in response to a devaluation.

The effect on wages is very different. The upward pressure caused by the devaluation (via increase in prices) and the productivity shock dominate during the first year and wages increase by 0.6 percent. However, effects of these shocks are eventually overwhelmed by the downward pressure caused by the increase in unemployment and the decline in capacity utilization. As a result, wages decline by 0.3 percent within a two year period.

A devaluation of the nominal effective exchange rate tends to increases the consumer prices. The other shocks have—ceteris paribus—opposite effects. Thus prices increase marginally in the beginning, but this trend is later reversed and the overall result is no change in the inflation rate in eight quarters from the beginning of the shock.

To summarize, the results of the simulations suggest that, in the early 1990s, the adverse shocks to the aggregate demand side may indeed have been strong enough to cause a recession of a sufficiently large magnitude to offset the inflationary effects of devaluation entirely. As the next section indicates, these results are consistent with the long-run experience.

2. Model performance

This section explores to what extent recent inflationary developments in Spain were in accord with the previous long-run experience. To answer this question, the actual past growth rates of endogenous variables are compared to the values resulting from the simultaneous simulation of the four dynamic equations of the model. This simulation was performed on 20 quarters—from the beginning of 1989 till the end of 1993. Chart 2 presents results of the quarterly inflation rates of the endogenous variables. Table 5 in Annex B summarizes quarterly and year-to-year results for 1991, 1992 and 1993.

Simulation to Evaluate Model Performance

(Quarterly percent changes)

Citation: IMF Working Papers 1996, 114; 10.5089/9781451943412.001.A001

Source: IMF, OECD Analytical Database.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Simulation to Evaluate Model Performance

(Quarterly percent changes)

Citation: IMF Working Papers 1996, 114; 10.5089/9781451943412.001.A001

Source: IMF, OECD Analytical Database.

• Download Figure
• Download figure as PowerPoint slide

Simulation to Evaluate Model Performance

(Quarterly percent changes)

Citation: IMF Working Papers 1996, 114; 10.5089/9781451943412.001.A001

Source: IMF, OECD Analytical Database.

• Download Figure
• Download figure as PowerPoint slide

The overall performance of the model is satisfactory, considering the fact that most of the estimated data points are within the one-standard-deviation range of the actual series. ^24/ The simulated values follow the actual ones reasonably closely, possibly with the exception of wages. This suggests that the model is a good representation of the long-run inflationary experience. The main implication of this result is that the relatively low inflation that followed the recent devaluation should not be regarded as surprising.

VII. Conclusions

This paper focuses on the inflationary impacts of the devaluations associated with the 1992-93 ERM crisis. A supply-side model of inflation determination is estimated on Spanish quarterly data. The determinants of the aggregate-demand side, such as fiscal and monetary policies or the foreign demand, are incorporated via proxies for the cyclical position of the economy. Channels of exchange rate pass-through are modeled together with domestic sources of inflation. The Johansen maximum-likelihood method is used to obtain long-run cointegrating vectors of import and export prices, wages and consumer prices. Further, application of the error-correction model yields dynamic equations to determine short-run inflationary behavior.

The estimation of the model suggests that Spain, is a price-taker for imported goods and its export prices are more heavily influenced by world prices than by domestic production costs. There is support for the pricing to market hypothesis, since the traded goods prices respond with a greater lag to exchange rate changes than to changes in world prices.

The analysis shows that real wages are highly positively correlated with productivity, possibly because of the strong bargaining power of the unions. However, although there is evidence of persistence in unemployment rates, there are indications of nominal wage rigidity. Wage inflation slowed down marginally after the entry of Spain in the ERM. In general, changes in the unemployment rate do not have a significant impact on the wage inflation, which shows the irresponsiveness of wage setting to labor market conditions.

Domestic costs dominate import prices in the determination of the general price level. Prices are relatively rigid; consumer price inflation responds to wage inflation only with a four-quarter lag. Overall, a slow speed of adjustment towards the long-run equilibrium is strongly indicated for all variables.

Simulations were performed on the estimated model: in particular, actual changes observed in 1992-93 the exchange rate and cyclical variables were imposed contemporaneously. Results indicate that the entire effect of the devaluation may have been eliminated by changes in other exogenous determinants of inflation. Since the model appears to fit well the overall historical behavior of Spanish inflation, the low inflation rate following the 1992-93 devaluations should not be considered an unusual phenomenon.