Gaining Policy Credibility for a Disinflation: Ireland’s Experience in the EMS
Author:
Jeroen J. M. Kremers https://isni.org/isni/0000000404811396 International Monetary Fund

Search for other papers by Jeroen J. M. Kremers in
Current site
Google Scholar
Close

Given a need for disinflation, the adoption of a (semi-)fixed exchange rale policy vis-à-vis a low-inflation country may provide a source of discipline, enhancing the disinflation’s credibility and reducing its detrimental impact on the economy. This paper presents empirical evidence that inflation expectations in Ireland were moderated by entry into the exchange rate mechanism of the EMS, indicating that Ireland’s disinflation was successful in deriving credibility from the exchange rate policy. A loss of competitiveness at an early stage of the disinflation appears to have been important in establishing its credibility.

Abstract

Given a need for disinflation, the adoption of a (semi-)fixed exchange rale policy vis-à-vis a low-inflation country may provide a source of discipline, enhancing the disinflation’s credibility and reducing its detrimental impact on the economy. This paper presents empirical evidence that inflation expectations in Ireland were moderated by entry into the exchange rate mechanism of the EMS, indicating that Ireland’s disinflation was successful in deriving credibility from the exchange rate policy. A loss of competitiveness at an early stage of the disinflation appears to have been important in establishing its credibility.

The desire to bring down inflation was an important motivation for inflation-prone countries such as Italy and Ireland to take part in the exchange rate mechanism (ERM) of the European Monetary System (EMS). It is often argued that, given a need for disinflation, participation in a mechanism of fixed but adjustable exchange rates with a group of Low-inflation countries can provide a source of discipline, enhancing the disinflation’s credibility and moderating its detrimental effects on the economy. Nevertheless, even though economic theory and public discussion have granted this role of policy credibility a degree of prominence, not much empirical research has been carried out to ascertain its practical relevance.1

Giavazzi and Giovannini (1989) conclude a theoretical review with the observation that, “if there is any advantage from pegging to a strong currency in a disinflation, [it] comes from a shift in price setters’ expectations, which makes the output loss of reducing inflation smaller.” For this argument to have empirical significance in the case of the ERM, two conditions must be satisfied. First, it must be true that the ERM has enhanced the credibility of disinflation—that is, that it has reduced expected inflation. Second, it must be true that the expectation of lower inflation has led to a moderation of wage settlements. In a situation where wages are indexed on actual inflation, participation in a strong-currency club may produce lower inflation through the impact on wages of adverse developments in competitiveness and employment, without however reducing the output cost of disinflation.

This paper examines whether disinflation policy in Ireland has derived credibility from its participation in the ERM. Ireland’s experience is of special interest in two respects. First, the Irish economy is relatively small and influences the rest of the ERM only to a limited extent. This factor facilitates the theoretical and empirical analysis. Second, Ireland’s experience is particularly suitable for inference because its entry into the ERM constituted a distinct change of exchange rate regime, more so than the entry of any of the other ERM participants. For over half a century the Irish pound had remained at a one-for-one, no-margins parity with sterling; not only did Ireland’s entry abruptly discontinue this parity, but, in addition, the United Kingdom decided not to participate in the ERM.

The analytical background of the paper is presented in Section I, which first summarizes theoretical arguments underlying the hypothesis that the ERM might help reduce the output cost of disinflation, and then briefly discusses how other studies have tested the empirical validity of this hypothesis. As the paper was motivated in part by a concern that some of the tests applied in the literature may produce unreliable inference. Section II sets out an alternative approach to modeling the credibility effect. Section III applies this approach to Irish data, producing strong evidence for the success of the authorities in deriving credibility for their disinflation from EMS membership. Section IV examines more closely the means and speed by which credibility was gained. Credibility alone, however, is not sufficient to mitigate the output cost of disinflation; in fact, Dornbusch (1989a) considers the Irish stabilization in this broader sense a failure. The credibility effect identified in this paper is placed in a wider perspective in Section V.

I. Analytical Background

At a theoretical level, some of the costs and benefits of ERM participation for a government that wishes to bring down inflation are analyzed by Giavazzi and Pagano (1988). In their model, the policymaker, on the one hand, is tempted to inflate so as to boost output, but, on the other, wishes to promote a low rate of inflation.2 The welfare gain of a credible participation in the ERM comes from the additional cost placed on inflation (loss of international competitiveness between realignments). Giavazzi and Pagano show that, given this policy configuration, the average rate of inflation of an inflation-prone participant will be lower than it would have been without the discipline of the ERM (but higher than that of its less inflation-prone partners) provided losses of competitiveness can be compensated by realignments at most to the extent necessary to ascertain purchasing power parity on average. Inflation differentials within the ERM will be related to the frequency of realignments; in the extreme case with no realignments, the balance of payments constraint would ultimately dictate full convergence of inflation (as under parity with sterling).3 This setup does not incorporate any strategic interactions between the investigated country and its ERM partners, which may be an acceptable simplifying assumption for a relatively small economy such as Ireland.

The analysis so far is conditional on the assumption that credibility is established immediately. But in practice to build a credible reputation for a new policy may take time and involve costs. To the extent that the inception of the ERM was accompanied by uncertainty about its robustness, and particularly about the prospects for frequent and sizable realignments, inflation expectations may have been slow to adjust to the equilibrium examined by Giavazzi and Pagano. An ensuing initial loss of competitiveness and, hence, of output, coutd have served to establish the new policy’s credibility.4

To evaluate the benefit of the ERM for inflation-prone participants, it is therefore necessary to assess its success in reducing inflationary expectations. Giavazzi and Giovannini (1988, 1989) address this issue by estimating vector autoregressions (VARs) for prices, wages, and output (conditional on lagged money and import prices) with data for the period preceding the ERM. Significant overprediction by this model of inflation during the ERM might have been consistent with a credibility effect, but, in fact, statistical tests cannot reject parameter stability between these subperiods for any of the countries investigated.5 At a less formal level, Giavazzi and Giovannini find that dynamic simulations of their VARs tend to overpredict the actual paths of inflation, but only starting several years after the inception of the EMS. They interpret this overprediction as a belated credibility effect, but acknowledge that the evidence is tenuous, since they do not compute the standard deviations of the prediction errors and hence are unable to determine whether the overprediction is significant.

Perhaps more important, concern can be raised with respect to the reliability of VARs as vehicles for inference regarding the stability of specific structural parameters: it is difficult to associate instability of the coefficients of a VAR with specific changes in structural parameters. In particular, Giavazzi and Giovannini (1988, Table 2) note that the VAR for Ireland starts to overpredict price inflation more than three years after the inception of the EMS, while overprediction of wage inflation begins already in 1980. Would this imply that the Irish authorities managed to generate a redundant credibility effect two years after successfully moderating wage settlements, or might it instead reflect other, unidentified features of the underlying model? According to the analysis in Sections III and IV, the latter appears to be true.

An approach more akin to that adopted below is taken by Artis and Ormerod (1987), who examine whether an autoregressive predictor of inflation in the Federal Republic of Germany contributes more to the explanation of inflation in other ERM countries after 1979 than before. The results are mixed, suggesting credibility effects in France and Italy, but not in Belgium and the Netherlands.

II. Specification of the Empirical Model

This paper hypothesizes that consumer price inflation in Ireland depends on two elements—namely, cost factors and international price competition. Cost factors may influence price inflation in the non-tradables sector, where international competitive pressure is not prominent (see, for example, Bruno (1978) and Honohan and Flynn (1986)). But the paper hypothesizes that the tradables sector acts largely as a price taker. In the presence of extensive international price arbitrage, highlighted by numerous authors,6 price expectations in this sector will be based predominantly on expected competitor prices (in Irish currency). Moreover, even in the nontradables sector Irish prices may remain related to those of foreign competitors if significant factor price arbitrage takes place and the productivity performance in Ireland is not dissimilar to that abroad.

Before Ireland joined the ERM, the Irish pound had been fixed at parity with sterling for over half a century, and no regulatory impediments had existed on capital movements between Ireland and the United Kingdom—a situation that rendered Ireland’s monetary policy and price behavior quite dependent on their respective counterparts in the United Kingdom.7 Hence, in a model of inflation expectations in Ireland one would anticipate substantial weight being assigned to expected U.K. inflation.

The subsequent termination of the parity with sterling was motivated by the policy goal of reducing Ireland’s inflation through stabilizing its exchange rate vis-à-vis the currencies of the low-inflation ERM countries.8 Should this policy have been credible, then one would anticipate a shift of weight in the inflation equation for Ireland away from expected U.K. inflation toward inflation expected for the rest of the ERM, with, presumably, unchanged weights on the cost variables.

In addition to being influenced by expected foreign inflation, expectations are likely to reflect feedback from international price competitiveness (the real effective exchange rate) if the exchange rate is—at least to some extent—credibly fixed. To the extent that the policymaker refuses to (say) devalue to accommodate excessive inflation, a loss in the level of competitiveness will have to be recouped by moderating domestic relative to foreign inflation, either directly through the influence of competitiveness on the process of wage formation, maintaining profitability intact, or indirectly through initially deteriorating profitability and employment placing a downward pressure on wages; the time lags of the direct channel are likely to be shorter than those of the indirect channel. Since the primary motivation for Ireland’s decision to join the ERM was the desire to terminate its participation in the inflation-depreciation spiral of the United Kingdom, a second sign of policy credibility in the ERM would be a stronger influence of competitiveness on expected inflation.

Another influence on inflation expectations could emanate from fiscal variables if indeed governments rely more extensively on money seigniorage as a source of revenue the higher the degree of public indebtedness and the higher the tax rates (for example, Grilli (1989) and Roubini and Sachs (1989)). However, two arguments lead one to discount the possibility that seigniorage has been actively used as a budgetary instrument in Ireland. First, during the pre-ERM long-term parity with sterling there was probably little scope for the Irish authorities to influence the revenue from seigniorage. Second, during the ERM the authorities kept almost half of the public debt denominated in foreign currency, and thus effectively forwent the option of monetization.9 Nevertheless, some tests for the possible influence of the level of public debt, deficits, and nonseigniorage revenue on expected inflation are discussed at the end of Section III.

The theoretical ingredients are brought together in the following expression for inflation expectations in Ireland:

π t * = x ¯ t * a ¯ 1 + z ¯ 1 t 1 a ¯ 2. ( 1 )

The variable πt, denotes Irish inflation; variables with asterisks denote expectations (based on information available at the end of the previous period); x¯t* is a (1 x 2) (row) vector of explanatory expectations variables (underlined symbols are vectors; x1t* equals inflation expected for the United Kingdom and x2t* equals inflation expected for the ERM excluding Ireland); z¯1t1 is a (1 x m) vector of lagged explanatory variables (cost factors and competitiveness);a¯1 and a¯2 are the corresponding (2 x 1) and (m x 1) coefficient vectors.

It is assumed in this formulation that in the nontradables sector pricing is implemented so that the cost factors can be entered with at least a short lag (one quarter in the estimates below). For the sake of a transparent notation, the lag structure is kept as simple as possible in this section, and fiscal variables are excluded.

Expectations are assumed to be based on an efficient use of available information:

π i * = E ( π t | z ¯ t 1 ) = π t u t

x ¯ i * = E ( x ¯ t | z ¯ 2 t 1 ) = x ¯ t v ¯ t . ( 2 )

The notation E(|) indicates the mathematical expectation of the first argument conditional on the information represented by the second argument. The (1 x (m + n)) vector z¯t1=(z¯1t1,z¯2t1) includes all the information that is used; its second component contains the information underlying x¯t* (the determination of x¯t* is specified in equation (3) below).10 The disturbance vector (ut,v¯t) is assumed to be distributed independently over time (with zero mean and constant nonsingular variance matrix); it is also independent of z¯t1.

The purpose of this paper is to conduct inference on the coefficients of equation (1); substituting equation (2) into equation (1), the latter can be expressed in terms of observable variables:

π t = h ¯ t a ¯ + w t , ( 1 )

where h¯t=(x¯t,z¯1t1),a¯=(a¯1´,a¯2), and wt=utv¯ta¯1. However, given equation (2), h¯t and wt are correlated, and therefore inference based on estimating equation (1´) by ordinary least squares (OLS) would be unreliable (the estimate of a¯ would be inconsistent). For consistency, account must be taken of the endogeneity of x¯t. For that purpose, assume that the expectations of x¯t are formed according to

E ( x ¯ t | z ¯ 2 t 1 ) = z ¯ 2 t 1 B , ( 3 )

where B is a (n x 2) coefficient matrix.

The model as it stands is of the type analyzed by Pesaran (1988, pp. 164-70), who discusses various estimation techniques. The most efficient approach would be to apply a full information method to the complete model, equations (1') and (3) jointly. However, the disadvantage of this approach would be its sensitivity to possible misspecification of equation (3). Limited information methods that do not rely on the full estimation of equation (3) are generally more robust to such misspecification (Pesaran (1988, p. 162)). Since the coefficients of present interest (a¯1) appear only in equation (1') and not in equation (3), a limited information method is justified. Specifically, a¯ will be estimated by the Instrumental Variables (IV) estimator:

a ¯ ^ = ( H P Q H ) 1 H P Q π ¯ , ( 4 )

where H = [X, Z1] is the (Tx(2 + m)) matrix of observations on h¯t (T is the sample size); PQ=Q(Q'Q)-1Q' is a (T x T) matrix projecting orthogonally onto the space spanned by the columns of the instrumental variables matrix Q = [Z1, Z3]; Z3 is the (T x q) matrix of observations on the instruments z¯3t1 (all uncorrected with wt) included in addition to z¯1t1 and π¯ is the vector of observed Irish inflation.

Three aspects of this IV estimator are noteworthy. First, under the usual assumptions the estimates are consistent and asymptotically efficient (Bowden and Turkington (1984, pp. 110-11)). In finite samples their efficiency will be enhanced by selecting z¯3t1 as closely correlated as possible with z¯2t1 (Bowden and Turkington (1984, par. 2.5)); for the identification of the coefficients of interest, the dimension of z¯3t1 must at least equal that of x¯t.

Second, a¯^ can be interpreted as a two-step estimator. In the first step, the explanatory variables of equation (1') are regressed (by OLS) on the instruments, producing fitted values Ĥ PQH. In the second step, the IV estimate of a¯ is derived by regressing π¯ on these fitted values:

a ¯ ^ = ( H ^ H ^ ) 1 H ^ π ¯ . ( 4 )

Using the fact that PQ is idempotent, it can easily be verified that the estimators tn equations (4) and (4') are identical. If, for expositional purposes, it were assumed that z¯1t1 was uncorrected with both x¯t and z¯3t1 then equation (4') would simplify to11

a ¯ ^ 1 = ( X ^ X ^ ) 1 X ^ π ¯

a ¯ ^ 2 = ( Z 1 Z 1 ) 1 Z 1 π ¯ , ( 4 )

where X^=(Z3Z3)1Z3X. This simplification would offer the intuitive interpretation that the IV estimate of a¯1 had been obtained first by regressing the endogenous explanatory variables on a set of appropriate instruments (Z3), and then by regressing Irish inflation on the fitted values from this first-stage regression. The underlying assumption is unlikely to be satisfied, however, and, since it is also unnecessary, the IV estimator (4) will be used below.

Third, it would be correct to interpret the first-stage fitted values X^ = PQX as expected values of X and the second-stage fitted values π¯^=PH˙π¯ as expected values of π¯ only if the instrument set Q included all the information underlying X* (that is, Z2). In practice it may be difficult to ascertain whether this condition is satisfied; evidence to the contrary might arise if diagnostic tests indicated that the regression of X on Q did not represent a proper model of expectations formed rationally (for example, the presence of residual autocorrelation could suggest that relevant information had been omitted). The attraction of the IV approach is that even under those circumstances it allows inference on the parameters of interest; in this paper the (marginal) models for U.K. and ERM inflation will not be scrutinized, nor will any attempt be made to interpret fitted values as expectations.

III. Empirical Results

In this section the model for inflation expectations expressed in terms of observable variables (equation (1')) is estimated with the IV method (4). The two explanatory expectations variables (inflation in the United Kingdom and inflation in the rest of the ERM) are treated as endogenous (denoted by an asterisk). The choice of additional instruments (z¯3t1 above) was based on some prior regressions for U.K. inflation (denoted in this section by ΔptUK)12 (this suggested the following instruments: U.K. inflation lagged by one quarter; U.K. import price inflation lagged by one quarter (Δiuvt1UK); world consumer price inflation lagged by four quarters (Δpt4W); oil price inflation lagged by five quarters (Δpt5OIL)); and on some prior regressions for ERM inflation (denoted in this section by ΔptERM); this suggested, as an additional instrument, ERM inflation lagged by one quarter.13. Data sources are given in Appendix I.

Two types of exogenous explanatory variables (z¯1t1 in the notation of Section II) were considered: lagged cost variables (wages, import prices, and oil prices) and lagged competitiveness (the real effective exchange rate). The precise lag structure was determined first by including a large number of lags (up to eight quarters) and then by selecting a specification with lags that were plausible from a theoretical perspective, significant, and satisfactory in the light of diagnostic tests (for residual autocorrelation, nonnormality, heteroscedasticity, and parameter instability; the reported tests are briefly explained inAppendix II). The role of fiscal variables is discussed at the end of the section.

Before proceeding to the empirical results, it may be helpful to comment briefly on the specification and interpretation of the feedback from competitiveness. In the extreme case of a fully accommodative exchange rate policy, there would be no such feedback, since competitiveness would be maintained by adjusting the nominal exchange rate. At the other extreme, with a rigidly fixed exchange rate, the entire burden of keeping competitiveness intact would fall on the domestic price level; shocks producing deviations of competitiveness from its equilibrium level would generate feedback onto domestic inflation. Hence, the lagged level of competitiveness, as well as, possibly, dynamic terms in the same variable (rates of change and acceleration) could play a significant role in a model of expected inflation. Feedback could operate relatively rapidly if wages were sensitive to competitiveness and adjustment took place while maintaining profitability, activity, and employment intact; a less direct and probably slower channel could operate through changes in profitability, activity, employment, and, ultimately, wages. Of course, a range of possible intermediate cases also exists in which both the exchange rate and inflation adjust. Based on these considerations, a credible shift from a relatively accommodative to a relatively rigid exchange rate policy would presumably be reflected in a stronger influence on expected inflation from the level of competitiveness (and also, perhaps, from dynamic terms in the same variable). Shorter lags could indicate a more responsive process of wage formation (and therefore reduced Output costs of disinflation).

The modeling procedure yielded the following preferred model for Irish inflation (denoted in this section by ΔptIR) before the ERM:14

Δ p t I R = 0.63 ( 0.09 ) Δ p t U K * + 0.21 ( 0.08 ) Δ p t E R M * + 0.14 Δ w t 1 + 0.019 ( 0.008 ) Δ p t 3 O I L

0.15 ( 0.07 ) Δ Δ ( r e e r t 2 + r e e r t 4 ) / 2 ( 5 )
article image

Estimated coefficient standard errors are in parentheses under the coefficient estimates, and 5 percent critical values are given behind the test statistics (see Appendix II); σ^ denotes the estimated equation standard error; w denotes the hourly wage rate; and reer denotes the real effective exchange rate of the Irish pound relative to the currencies of 16 industrial trading partners (Irish relative to foreign consumer price level, corrected for exchange rate changes). In regression (5), and in all the regressions that follow, the inflation homogeneity restriction could not be rejected and was subsequently imposed. This restriction is common in the literature on wage-price dynamics (for example, Bruno (1978, 1980); Grubb, Jackman, and Layard (1982, 1983); Artis and Ormerod (1987); and Giavazzi and Giovannini (1988, 1989)); it ascertains that in a hypothetical steady-state growth equilibrium all price variables increase at the same rate. In regression (5), therefore, the first four coefficients sum to unity, and. reflecting the imposed restriction, one of the coefficients is reported without an estimated standard error.15

Several features of regression (5) are noteworthy. First, it appears possible to estimate a parsimonious model for inflation expectations in the pre-ERM period specified along the lines discussed in Section II. All the diagnostic tests are satisfied at a significance level of 5 percent (test failures at 5 percent are marked by asterisks). Second, the influence of expected U.K. inflation indeed appears to be dominant during this period, with an estimated coefficient three times as large as the estimated coefficient on expected ERM inflation. That the latter is significant at all probably reflects some degree of fixity of sterling relative to other ERM currencies during part of the pre-ERM period (Bretton Woods and the “snake”). Third, the influence of cost factors (wages and oil prices16) appears to be relatively small. Fourth, there was no significant feedback from lags (up to eight quarters) of the level of competitiveness or changes thereof; only sharp movements (accelerations) in competitiveness were found to have a discernible impact on inflation expectations in pre-ERM Ireland. This finding suggests that competitiveness was not expected to be maintained by adjustments in domestic inflation, but rather by the accommodative stance of the (U.K.) exchange rate policy. Finally, all of the following additional variables appeared to be insignificant throughout the analysis: lagged values of Irish, U.K., and ERM consumer prices; lagged Irish import prices and world consumer prices; and further lags of wages and oil prices.

This model breaks down when extended beyond 1979. The estimate of regression (5) for the entire sample (1965:1-1986:4) fails almost all of the diagnostic tests:

Δ p t I R = 0.72 ( 0.11 ) Δ p t U K * + 0.06 ( 0.10 ) Δ p t E R M * + 0.19 Δ w t 1 + 0.024 ( 0.009 ) Δ p t 3 O I L 0.09 ( 0.05 ) Δ Δ ( r e e r t 2 + r e e r t 4 ) / 2 ( 6 )
article image

This regression suffers residual autocorrelation, nonnormality, and heteroscedasticity; it fails the forecasting test for the ERM period; the instruments are correlated with the residuals; and the estimated residual variance for the total sample is nearly double that estimated for the pre-ERM period. The next step in the investigation is to find the source of this model breakdown.

The first issue to be addressed more precisely is whether the timing of the breakdown is related to Ireland’s entry into the ERM in the first quarter of 1979. For that purpose, regression (5) is re-estimated by recursive IV, starting from the sample 1965:1-1972:4 and successively extending the sample by a single observation until the estimate for the total period 1965:1-1986:4 (that is, model (6)) is reached. The resulting sequence of estimated equation standard errors and the corresponding one-step residuals17 are depicted in Figure 1, It turns out that the rise in σ^ occurred soon after the inception of the ERM, in association with large positive one-step residuals in 1979:1 and from 1980:2 through 1982:2.

Figure 1.
Figure 1.

Recursive Estimation of the Pre-ERM Expectations Model

Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A004

Source: Equations (5) and (6).Note: The figure depicts the sequence of estimated equation standard errors and one-step residuals (see text footnote 17) of the pre-ERM model, obtained by estimating equation (5) by recursive instrumental variables, successively extending the sample by a single observation from 1965:1-1972:1 to 1965:1-1986:4. The estimate for 1965:1-1978:4 is given in equation (5), and that for 1965:1-1986:4 in equation (6).

The positive sign of most of the one-step residuals during the first few years of the ERM indicates that inflation in Ireland initially was higher than predicted on the basis of the pre-ERM model. This result may have reflected, in part, uncertainty about the new exchange rate policy: “During 1979 strenuous attempts were made to obtain an agreement on an incomes policy based on the expectation of inflation falling to single figures. The failure to obtain such an agreement was a clear indicator that expectations about inflation had not been radically changed as a result of EMS entry” (Walsh (1983, p. 173)). Compounding the initial uncertainty, the Irish pound depreciated by almost 15 percent in nominal effective terms during the first two years of the ERM, reflecting the stabilization of its rate within the ERM in combination with the strengthening of sterling (Figure 2). Rather than confirming the expectation that Ireland had switched to a harder currency peg, this development reinforced the inflationary spiral (Figure 3 and Walsh (1983)), while at the same time allowing Ireland to record a modest gain in competitiveness (Figure 2).

Figure 2.
Figure 2.

Exchange Rate Developments

(Cumulative changes in percent; 1979:1 =0)

Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A004

Source: Appendix I.Note: Quarterly data; foreign currency per Irish pound.a Weighted average of ERM partners.b Weighted average of 16 industrial partners, including ERM.c Based on consumer price indices; a rise indicates a real appreciation of the Irish pound.
Figure 3.
Figure 3.

Inflation

(In percent)

Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A004

Source: Appendix I.Note: Quarterly consumer price index; rate of increase relative to the corresponding quarter of the previous year.

While the depreciation and the upsurge of inflation seemed to represent a setback for the authorities’ disinflation policy, it should, as noted in Section I, perhaps not have come as a surprise that the establishment of credibility for the new policy took time and effort. Before focusing on the possible accrual of credibility (in Section IV), it is first assessed on the basis of the entire ERM sample whether any credibility was ultimately gained at all. The evidence is affirmative. Equation (7) reports an estimate of the expectations model for the entire sample, allowing shifts in the coefficients on U.K. and ERM inflation (as well as their instruments) and on competitiveness at the beginning of the ERM, while retaining the (tested) homogeneity restraint in each of the two subperiods.18

Δ p t I R = [ 0.64 ( 0.09 ) Δ p t U K * + 0.20 ( 0.07 ) Δ p t E R M * + 0.14 ( 0.06 ) Δ w t 1 0.14 Δ Δ ( r e e r t 2 + r e e r t 4 ) / 2 ] D 6578

+ [ 0.70 ( 0.08 ) Δ p t E R M * + 0.28 Δ w t 1 0.20 ( 0.04 ) Δ Δ r e e r t 1 0.11 ( 0.02 ) r e e r t 2 + 0.51 ( 0.11 ) ] D 7986

+ 0.017 ( 0.007 ) Δ p t 3 O I L ( 7 )
article image

The variables D6578 and D7986 represent multiplicative dummies that take the value of unity for the subperiods 1965:1-1978:4 and 1979:1-1986:4, respectively, and zero elsewhere. Equation (7) satisfies ail of the diagnostic tests, and, in contrast to equation (6), its estimated equation standard error is virtually identical to that estimated for the pre-ERM period (see equation (5)). The constant term was added for the ERM period to allow for the nonzero sample mean of the lagged level of competitiveness.

The evidence indicating the presence of a credibility effect is twofold. First, there is a sharp shift of weight from the coefficient on expected U.K. inflation to that on expected ERM inflation: more general regressions showed that the former became small and insignificant after 1979 (even when estimated for a very brief subperiod starting immediately in 1979:1). and therefore it was dropped. Second, during the ERM period the lagged level of competitiveness became an important determinant of expected inflation, with losses of competitiveness expected to be recouped by adjustments in Ireland’s inflation. Taken together, these findings provide strong evidence that expectations did indeed adjust to the new exchange rate policy within the EMS. It is also interesting to note that the lag on the acceleration of competitiveness became half a year shorter, which may reflect the more direct sensitivity of wages to competitiveness, as discussed above.

Given the constancy of the estimated coefficients (discussed in Section IV), inspection of equation (7) suggests that the following two restrictions may be valid: (1) the coefficient on expected U.K. inflation and the coefficient on expected ERM inflation, both before the ERM, may sum to the coefficient on expected ERM inflation after 1979 (implying, given equation (7) that the coefficient on lagged wage inflation before the ERM equals the corresponding coefficient after 1979); and (2) the coefficient on the variable for the acceleration of lagged competitiveness before the ERM may equal the corresponding coefficient after 1979. Because a joint significance test cannot reject these restrictions, they are imposed so as to arrive at a parsimonious preferred model for Irish inflation expectations before and during the ERM:

Δ p t I R = [ 0.59 ( 0.09 ) Δ p t U K * + 0.22 Δ p t E R M * 0.19 ( 0.03 ) Δ Δ ( r e e r t 2 + r e e r t 4 ) / 2 ] × D 6578

+ [ 0.81 ( 0.04 ) Δ p t E R M * 0.19 Δ Δ r e e r t 1 0.11 ( 0.02 ) r e e r t 2 + 0.51 ( 0.11 ) ] D 7986 + 0.17 Δ w t 1

+ 0.020 ( 0.007 ) Δ p t 3 O I L ( 8 )
article image

It should be pointed out that the validity of the expectations interpretation given to the current model is not affected by whether, in about 1979, a regime shift occurred in the process generating U.K. inflation. Such a possible regime shift might affect the correlation between the instruments used and (endogenous) U.K. inflation, but considering that the instruments are lagged and no residual autocorrelation is apparent in the estimated equations for Irish inflation, the validity of the current approach would not be impaired.19 A formal treatment of these issues can be found in Hendry (1988).

The possibility that fiscal variables influence inflation expectations was examined by testing for the significance of the public debt and nonseigniorage government revenue (both in relation to gross national product (GNP)) in models (7) and (8). Since it is not clear a priori whether lagged or expected values of these variables might be relevant, two alternative specifications were tested: (i) the addition of lagged values (first through fourth order) of both of these ratios as exogenous variables; and (2) the addition of the current values of both of these ratios as endogenous variables as well as the addition of their lagged values (first through fourth order) as exogenous variables.20 To allow for differences in the possible influence of these variables on inflation expectations before and during the ERM, all tests were specified allowing for coefficient shifts in 1979:1. The tests unambiguously reject the significance of the fiscal variables for each of the two subperiods and in each of the two alternative specifications. The same is true for a variety of more specific tests, adding individually either of the two variables in either of the two subperiods (in either of the two specifications described above). Given that the time pattern of the fiscal deficit relative to GNP is very similar to that of the first difference of the debt/GNP ratio, and that adding several lags of the debt/GNP ratio encompasses adding lags of the first difference as a special case, these results suggest that the levels of public debt, deficits, and nonseigniorage revenue relative to GNP have not had a simple systematic impact on inflation expectations in Ireland.21

IV. The Process of Gaining Credibility

To examine more closely the accrual of credibility for Ireland’s disinflation in the ERM, two aspects of credibility are distinguished: the public’s estimate of the parameters underlying the new policy regime, and the degree of uncertainty attached to that estimate. The process of gaining credibility thus has two dimensions: the policymaker must make the public aware of the new policy, and reduce the uncertainty that may initially surround it.

How this can be done depends on the policymaker’s reputation. A policymaker with an impeccable reputation for reliability may need do no more than announce the new policy; even if it is an arduous one, such as a disinflation, the public may believe the announcement without reservation. If the policymaker’s reputation is not all that strong, the public may accept the announced policy as its best guess, but attach considerable uncertainty to it. If the policymaker is even less credible, the public may expect policy outcomes that are significantly different from the announcement. In the latter two cases detrimental economic effects may occur; in a disinflation, a lack of policy credibility may undermine the public’s willingness to moderate wage settlements. The question then arises as to how uncertainty can be reduced and credibility gained. One strategy could be just to stick to the announced policy and let uncertainty subside as time passes. However, it might be preferable to speed up this process by addressing the source of uncertainty more directly.

In the theoretical model of Vickers (1986), which is particularly relevant in the present context, the public is uncertain whether the monetary policymaker is a “dry” (with low tolerance for inflation) or a “wet” (with high tolerance for inflation). The uncertainty arises from the wet’s incentive to pose initially as a dry, raising the scope for an output-boosting inflationary shock sometime in the future. If in those circumstances the cost of initially generating a recession to gain credibility was smaller than the output cost of a long-drawn lack of credibility, it would be profitable for a dry government initially to act dryer than a wet government would be prepared to, thereby revealing its identity and dissolving the uncertainty.

Therefore, to be able to reap the benefits of the ERM, the Irish authorities may first have tried to build up a certain level of credibility by adopting a tough stance on the exchange rate and by tolerating an initial loss of competitiveness. Only when that level had been reached would the model of Giavazzi and Pagano (1988) become relevant; in that model (discussed in Section I) there is no need for competitive losses (other than temporarily between realignments) once credibility has been established.

This scenario appears to fit the data rather well: Figure 2 shows that the nominal effective exchange rate of the Irish pound relative to the other ERM currencies was kept constant during the first four years of the system, even though inflation was much higher in Ireland than elsewhere. This produced a steady pattern of real effective appreciation relative to the rest of the ERM, totaling almost 30 percent. However, the policy was ineffective initially owing to the strength of sterling; Ireland’s exchange rate constraint thus in fact became less binding in 1979-80, but the subsequent stability of the nominal effective exchange rate (reflecting a weakening of sterling offset by a strengthening of the U.S. dollar), in combination with Ireland’s relatively rapid inflation, resulted in a real effective appreciation of about 15 percent in 1981-82; this sharp loss of competitiveness was followed by several years of real exchange rate stability.22

Have these exchange rate developments helped to inform the public of the disinflation policy and to reduce the uncertainty that may have surrounded it at the outset? Some light may be shed on this issue by re-estimating model (7) by recursive instrumental variables, starting with a minimal number of observations needed for estimating the post-1979 coefficients.23 The interest of this exercise is twofold. First, it offers an assessment of whether the coefficient estimates of equation (7) are stable and, hence, whether the model is well specified. Second, the pattern of the standard errors associated with the coefficient estimates may yield insight into the process by which the uncertainty initially surrounding the disinflation was reduced. It is of course not surprising that the standard errors decline as more observations are added to the sample; this fall simply reflects that, as information accrues, the precision of the estimates improves. However, the interest lies in the pattern by which the standard errors fall. Relatively large drops occur during episodes when, in a statistical sense, the data are relatively “variable” (Campos and Ericsson (1988)). In the present context, this can be interpreted to mean economically that during such episodes a relatively large amount of information accrues, or that, in other words, uncertainty about the new exchange rate and inflation regime then abates and policy credibility is gained.24

The resulting time patterns of estimated coefficients and standard errors pertinent to the credibility effect are shown in Figures 4 and 5. Three features are noteworthy. First, there is no evidence of significant parameter instability (at no point do the estimates of either of the two coefficients move out of any of the previous 95 percent confidence intervals). Second, even before the real effective appreciation of the Irish pound in 1981-82 the estimated coefficient on expected ERM inflation was significantly larger than that for the pre-ERM period (Figure 4), suggesting that the Government’s intention to disinffate was perceived relatively early on. Third, by the end of 1980, after two years’ experience in the ERM, the estimated coefficient on lagged competitiveness was not yet significantly different from zero (Figure 5). But in the course of 1981-82, as Ireland incurred a sharp loss of competitiveness, the estimated coefficient on this variable became significantly negative.25 As shown in the bottom panels of Figures 4 and 5, the standard errors of both of the variables associated with the credibility effect fell rapidly in 1981-82. Only a moderate further reduction of uncertainty appears to have occurred after 1982. Hence, the process of gaining credibility seems to have taken place in various stages during 1979-82; both the entry into the ERM in 1979 and the competitive loss incurred in 1981-82 seem to have served an important purpose as a signaling device.

Figure 4.
Figure 4.

Recursive Estimation of the Coefficient on Expected ERM Inflation

Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A004

Source: Equation (7).Note: The figure depicts the sequence of estimates of the coefficient on expected ERM inflation, obtained by estimating equation (7) by recursive instrumental variables, successively extending the sample by a single observation from 1965:1-1980:4 to 1965:1-1986:4.
Figure 5.
Figure 5.

Recursive Estimation of the Coefficient on Lagged Competitiveness

Citation: IMF Staff Papers 1990, 001; 10.5089/9781451956863.024.A004

Source: Equation (7).Note: The figure depicts the sequence of estimates of the coefficient on lagged competitiveness, obtained by estimating equation (7) by recursive instrumental variables, successively extending the sample by a single observation from 1965:1-1980:4 to 1965:1-1986:4.

V. Conclusions

This paper presents evidence that Ireland’s disinflation policy has derived credibility from its participation in the exchange rate mechanism of the EMS. Before 1979. Irish inflation expectations mainly followed the expected movements of prices in the United Kingdom, and the influence of international price competitiveness on expected inflation was minor. In contrast, upon entry into the ERM, those expectations soon reflected the expected price behavior of ERM partners, and competitiveness became an important determinant of expected inflation in Ireland. The process of gaining credibility seems to have taken place in various stages during 1979-82, rather than several years after the start of the ERM, as argued by Giavazzi and Giovannini (1988); both the entry into the ERM and the competitive loss incurred in 1981-82 appear to have played an important role.

In a broader perspective, the Irish disinflation formed part of a stabilisation policy that also included a program of fiscal consolidation, limiting the scope for using demand management to alleviate the output cost of gaining credibility for the disinflation. The fiscal contraction under a fixed exchange rate would ideally have been accompanied by a moderation in wages, so that the reduction in expenditure brought about by fiscal consolidation could have been matched by expenditure switching (toward exports and import substitution) brought about by a gain in competitiveness. Dornbusch (1989a) considers the Irish stabilization in this broader sense a failure, mainly because the labor market did not deliver the competitive gain necessary to induce expenditure switching. As a result, employment in manufacturing fell at an average annual rate of about 2 percent during 1979-87, and unemployment rose from 8 percent of the labor force in 1978 to over 18 percent in 1987, while at the same time emigration increased. Moreover, monetary policy had to be tightened at times to defend the exchange rate in the face of large private capital outflows reflecting slippages in fiscal consolidation. The need to maintain short-term interest rates at a high level in turn aggravated the fiscal difficulties and further inhibited economic growth, leading to an increase in the ratio of government debt to GNP from 80 percent in 1978 to over 130 percent in 1987.26

The findings of this paper give rise to two footnotes to such an assessment. First, considering Ireland’s inflation history, it is not surprising that perfect credibility for the disinflation was not forthcoming immediately upon entry into the ERM. But this does not imply that participation in the ERM failed to help reduce the output cost of disinflation. Without the ERM, the performance of the Irish economy during the disinflation might have been even worse. The evidence of this paper indicates that the ERM did help to influence inflation expectations downward, which is consistent with the decline in long-term interest rates noted by Dornbusch. The question remains whether this in turn helped to moderate wage settlements; while more detailed research is required, the breakdown of the vector auto regression for Irish wage inflation reported by Giavazzi and Giovannini (1988) suggests that the wage formation process did indeed change after the inception of the ERM.

Second, Dornbusch (1989a, p, 209) concludes with the recommendation that “[t]he lesson for stabilization policy is clearly that governments must pay far more attention to the need for crowding-in through increased competitiveness. A major real depreciation at the outset of the program would provide the required offset to budget cutting.” The findings of this paper cast doubt on the feasibility of the second part of the recommendation in a situation where policy credibility is lacking. The Irish pound did in fact enter the EMS at a relatively favorable central rate (van Ypersele (1985)), and, in addition, depreciated by almost 15 percent in nominal effective terms within two years. However, an initial lack of policy credibility doomed to failure strenuous attempts at implementing an incomes policy designed to make crowding-in possible through increased competitiveness (Walsh 1983, p. 173));27 the nominal depreciation did no more therefore than produce accelerating inflation and thus a more arduous disinflation. Only after the real appreciation of 1981-82 did the disinflation get under way.

Nevertheless, it has subsequently taken a long time to rebuild competitiveness and reap the fruits of credibility.28 The Irish economy seems to have turned around only very recently; since 1987 fiscal adjustment has firmed, private capital outflows have stopped, wages have moderated, competitiveness has improved, exports and investment have picked up, and unemployment has started to fail (Bacon (1988), McAleese (1989), and Organization for Economic Cooperation and Development (1989)). Further research could therefore examine the factors (particularly fiscal, labor market, and industrial policies) that determine how credibility, once gained, can best be exploited to minimize the output costs of disinflation.

APPENDIX I Data Sources

Data sources are given below (variables in natural logarithms); data that are not available in published form can be obtained from the author.

article image
article image

The tests for the significance of fiscal variables discussed in Section III relied on annual data for the central government’s gross debt (1960-77: IFS, line 88; 1978-86: Budget 1989, p. 113); the central government’s nonseigniorage revenue (IFS, line 81); and nominal GNP (IFS, line 99a); the annual series representing ratios based on these data were subsequently interpolated to quarterly series.

APPENDIX II Test Statistics

This Appendix contains brief descriptions of the reported diagnostic tests, with degrees of freedom in brackets. Further background on these tests can be found in Hendry (1989). In the text, each reported statistic is followed (in parentheses) by its 5 percent critical value.

article image
article image

REFERENCES

  • Artis, M.J., and P. Ormerod, “Converging on the German Standard: Wage- Price Processes in Western Europe” (unpublished; England, University of Manchester, 1987).

    • Search Google Scholar
    • Export Citation
  • Bacon, Peter, “The European Monetary System: Sterling and the Irish Pound,” IrishBankingReview (Autumn 1988).

  • Baxter, Marianne, “The Role of Expectations in Stabilization Policy,” JournalofMonetaryEconomics, Vol. 15 (May 1985).

  • Bowden, Roger J., and Darrell A. Turkington, Instrumental Variables (Cambridge, England: Cambridge University Press, 1984).

  • Browne, F.X., “A Monthly Money Market Model for Ireland in the EMS,” Central Bank of Ireland, Annual Report, 1986.

  • Bruno, Michael, “Exchange Rates, Import Costs, and Wage-Price Dynamics,” JournalofPoliticalEconomy, Vol. 86 (1978).

  • Bruno, Michael, “Import Prices and Stagflation in the Industrial Countries: A Cross- Section Analysis,” The Economic Journal, Vol. 90 (September 1980).

    • Search Google Scholar
    • Export Citation
  • Campos, Julia, and Neil R. Ericsson, “Econometric Modeling of Consumers’ Expenditure in Venezuela,” International Finance Discussion Paper 325 (Washington: Board of Governors of the Federal Reserve System, June 1988).

    • Search Google Scholar
    • Export Citation
  • Dornbusch, Rudiger, “The European Monetary System, the Dollar and the Yen,” in The European Monetary System, ed. by F. Giavazzi, S. Micossi, and M. Miller (Cambridge, England: Cambridge University Press, 1988).

    • Search Google Scholar
    • Export Citation
  • Dornbusch, Rudiger, (1989a), “Credibility, Debt and Unemployment: Ireland’s Failed Stabilization,” Economic Policy, No. 8 (April 1989).

    • Search Google Scholar
    • Export Citation
  • Dornbusch, Rudiger, (1989b), “Discussion of ‘Seigniorage in Europe’ by V. Grilli,” in AEuropean Central Bank? Perspectives on Monetary Unification After TenYears of the EMS, ed. by M. De Cecco and A. Giovannini (Cambridge, England: Cambridge University Press, 1989).

    • Search Google Scholar
    • Export Citation
  • Driffill, John, “Macroeconomic Policy Games with Incomplete Information: A Survey,” European Economic Review, Vol. 32 (March 1988).

  • Engle, Robert F., “Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation,” Econometrica, Vol. 50 (July 1982).

    • Search Google Scholar
    • Export Citation
  • Giavazzi, Francesco, and Alberto Giovannini, “Can the European Monetary System Be Copied Outside Europe? Lessons from Ten Years of Monetary Policy Coordination in Europe,” NBER Working Paper 2786 (Cambridge, Massachusetts: National Bureau of Economic Research, December 1988).

    • Search Google Scholar
    • Export Citation
  • Giavazzi, Francesco, and Alberto Giovannini, “Interpreting the European Disinflation: The Role of the Exchange Rate Regime,” in Limiting Exchange Rate Flexibility: The EuropeanMonetary System, ed. by Francesco Giavazzi and Alberto Giovannini (Cambridge, Massachusetts: MIT Press, 1989).

    • Search Google Scholar
    • Export Citation
  • Giavazzi, Francesco, and Marco Pagano, “The Advantage of Tying One’s Hands: EMS Discipline and Central Bank Credibility,” European EconomicReview, Vol. 32 (1988).

    • Search Google Scholar
    • Export Citation
  • Godfrey, L.G., “Testing Against General Autoregressive Moving Average Error Models when the Regressors Include Lagged Dependent Variables,” Econometrica, Vol. 46 (November 1978).

    • Search Google Scholar
    • Export Citation
  • Grilli, Vittorio, “Seigniorage in Europe,” in A European Central Bank? Perspectiveson Monetary Unification After Ten Years of the EMS, ed. by M. De Cecco and A. Giovannini (Cambridge, England: Cambridge University Press, 1989).

    • Search Google Scholar
    • Export Citation
  • Grubb, D., R. Jackman, and R. Layard, “Causes of the Current Stagflation,” Review of Economic Studies, Vol. 49 (1982).

  • Grubb, D., R. Jackman, and R. Layard, “Wage Rigidity and Unemployment in OECD Countries,” EuropeanEconomic Review, Vol. 21 (March/April 1983).

    • Search Google Scholar
    • Export Citation
  • Hendry, David F., “The Encompassing Implications of Feedback Versus Feedforward Mechanisms in Econometrics,” Oxford Economic Papers, Vol. 40 (March 1988).

    • Search Google Scholar
    • Export Citation
  • Hendry, David F., “The Econometrics of PC-GIVE,” in PC-GIVE: An Interactive Menu-Driven Econometric Modelling Program (Oxford: University of Oxford, 1989).

    • Search Google Scholar
    • Export Citation
  • Honohan, P., and J. Flynn, “Irish Inflation in EMS,” Economic and SocialReview, Vol. 17 (April 1986).

  • International Monetary Fund, International Financial Statistics (Washington: International Monetary Fund, various years).

  • International Monetary Fund, World Economic Outlook (Washington: International Monetary Fund, October 1987).

  • Ireland, Central Statistical Office, Irish Statistical Bulletin (Dublin), March 1976.

  • Ireland, Ministry of Finance, Budget 1989 (Dublin: The Stationery Office, January 1989).

  • Jarque, CM., and A.K. Bera, “Efficient Tests for Normality, Homoscedasticity, and Serial Independence of Regression Residuals,” Economics Letters, Vol. 6 (1980).

    • Search Google Scholar
    • Export Citation
  • Leddin, A., “Portfolio Equilibrium and Monetary Policy in Ireland,” The Economicand Social Review, Vol. 17 (January 1986).

  • Mayer, T., “The EMS and External Imbalances of EMS Countries: Is There a Connection?” (unpublished; Washington: International Monetary Fund, February 1989).

    • Search Google Scholar
    • Export Citation
  • McAleese, D., “European Integration and the Irish Economy,” Public Administration, Vol. 35 (1987).

  • McAleese, D., , “Ireland—The Economy Is Doing So Well, There Is Talk of an Irish Miracle,” EuropeMagazine of the European Community (January/February 1989).

    • Search Google Scholar
    • Export Citation
  • McCormack, D., “Policy-Making in a Small Open Economy: Some Aspects of Irish Experience,” Central Bank of Ireland, Quarterly Bulletin (Winter 1979).

    • Search Google Scholar
    • Export Citation
  • McGuirk, Anne K., “Measuring Price Competitiveness for Industrial Country Trade in Manufactures,” IMF Working Paper 87/34 (Washington: International Monetary Fund, 1987).

    • Search Google Scholar
    • Export Citation
  • Murray, C.H. (1979a), “The European Monetary System: Implications for Ireland,” Central Bank of Ireland, Annual Report, 1979.

  • Murray, C.H., (1979b), “Living with the EMS,” Central Bank of Ireland, QuarterlyBulletin (Winter 1979).

  • Organization for Economic Cooperation and Development, EconomicSurveys—Ireland (Paris: OECD, 1989).

  • Pagan, A., “Econometric Issues in the Analysis of Regressions with Generated Regressors,” International Economic Review, Vol. 25 (February 1984).

    • Search Google Scholar
    • Export Citation
  • Persson, Torsten, “Credibility of Macroeconomic Policy: An Introduction and a Broad Survey,” European Economic Review, Vol. 32 (March 1988).

    • Search Google Scholar
    • Export Citation
  • Persson, Torsten, and Sweder van Wijnbergen, “Signalling, Wage Controls and Monetary Disinflation” (unpublished; Stockholm: Institute for ‘International Economic Studies, October 1988).

    • Search Google Scholar
    • Export Citation
  • Pesaran, M.H., The Limits to Rational Expectations (New York: Blackwell, 1988).

  • Roubini, Nouriel, and Jeffrey D. Sachs, “Political and Economic Determinants of Budget Deficits in the Industrial Democracies,” European EconomicReview, Vol. 33 (May 1989).

    • Search Google Scholar
    • Export Citation
  • Sargan, J.D., “Wages and Prices in the United Kingdom: A Study in Econometric Methodology,” in Econometric Analysis for National EconomicPlanning, ed. by P.E. Hart, G. Mills, and J.K. Whitaker (London: Butterworth, 1964).

    • Search Google Scholar
    • Export Citation
  • van Ypersele, Jacques, The European Monetary System: Origins, Operation andOutlook (Brussels: Commission of the European Communities, 1985).

    • Search Google Scholar
    • Export Citation
  • Vickers, J., “Signalling in a Model of Monetary Policy with Incomplete Information,” Oxford Economic Papers, Vol. 38 (November 1986).

    • Search Google Scholar
    • Export Citation
  • Walsh, B.M., “Ireland in the European Monetary System: The Effects of a Change in Exchange Rate Regime,” in International Economic Adjustment:Small Countries and the European Monetary System, ed. by M. De Cecco (Oxford: Blackwell, 1983).

    • Search Google Scholar
    • Export Citation
  • White, Halbert, “A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity,” Econometrica, Vol. 48 (May 1980).

    • Search Google Scholar
    • Export Citation
*

Jeroen J.M. Kremers, an economist in the European Department when this paper was written, is now with the Ministry of Finance in the Netherlands. He holds a D.Phil, from Oxford University, The paper benefited from comments by N. Ericsson, P. Honohan, D. McAleese, colleagues in the Fund, the referees, and seminar participants at Tilburg University. Some of the data were kindly provided by B. Smith of the Ministry of Finance in Ireland.

1

See Driffill (1988) and Persson (1988) for reviews of the literature.

2

The latter objective carries less weight, however, to the extent that the public budget relies on revenue from money seigniorage. By participating in the exchange rate mechanism (ERM), a country forgoes, to some extent, the liberty of determining the rate at which its government may collect seigniorage. Dornbusch (1988) argues that several high-inflation ERM participants underestimated the budgetary consequences of limiting this source of revenue. Nevertheless, at the time of entry into the ERM, the Governor of the Central Bank of Ireland (Murray (1979b)) attached great importance to control over the public finances.

3

This proposition abstracts from relative price movements.

4

Driffill (1988) reviews the theoretical literature on the need to generate recession to establish a policy reputation. “The Irish policymakers were well aware of such startup costs: “Given our past record, and the inflationary expectations which are now a part of our way of life, price stability, without pain, is as unattainable as taxation without tears” (Murray (1979b, p. 66)).

5

Denmark, France, the Federal Republic of Germany, Ireland, Italy, and the nonparticipating United Kingdom. The model for France seems to exhibit a significant parameter shift in 1979, but the presence of residual autocorrelation reveals more pervasive misspecification.

6

For example, McCormack (1979) and Walsh (1983).

7

The lack of autonomy in Irish monetary policy during this period is documented in Leddin (1986); Honohan and Flynn (1986) provide evidence (and further references) on the relation between Irish and foreign inflation.

8

See McCormack (1979), Murray (1979a,b) (Murray was Central Bank Governor), and Walsh (1983). Other factors included the falling U.K. share in Ireland’s external trade and the prospect of resource transfers from ERM partners (McAleese (1987)).

9

The total gross debt of the central government was equivalent to 80 percent of gross national product (GNP) just before the start of the ERM in 1978; it subsequently rose to more than 130 percent of GNP in 1988 (See Ireland (1989, p. 113)).

10

For the exposition, z¯1t1 and z¯2t1 are assumed not to overlap, but this assumption is relaxed in the empirical application below.

11

See also Pagan (1984, Theorem 4).

12

All lowercase variables are in natural logarithms.

13

Given that none of the data are seasonally adjusted, it appeared necessary also to include one seasonal dummy as an instrument for the entire period (unity in the third quarter and minus unity in the fourth) and one seasonal dummy as an explanatory variable for the ERM period (unity in the second quarter).

14

The empirical results of this paper were obtained with PC-GIVE by D. F. Hendry and the Oxford Institute of Economics and Statistics; see Hendry (1989). All the regressions also include two unreported dummies associated with a sharp discontinuity in the Irish consumer price statistics in 1975-76. The first dummy is unity in 1975:1 and minus unity in 1975:2, and the second is unity in 1975:3 and minus one half both in 1976:1 and 1976:2 (both dummies are data based and specified so that they have a mean of zero). See Appendix I and Honohan and Flynn (1986) on the need for these dummies.

15

In the estimation, one of the explanatory inflation variables was deducted from the dependent variable and from each of the other explanatory inflation variables, and was also, of course, included as an instrument.

16

In Irish pounds. As noted in the text, current and lagged values of Ireland’s overall import unit value index were irrelevant throughout.

17

The sequence of one-step residuals consists of the last residual of each of the successive regressions. If they are to be normally distributed with zero mean, they must be close to zero and about 95 percent of them must lie within the band delimited by two estimated equation standard errors.

18

The ERM came into operation in March 1979 following a resolution of the European Council on December 5,1978. The coefficient shift in the expectations model is therefore placed immediately after 1978:4. Though not made explicit in equation (7), the lag of the oil price variable was raised to four quarters after 1979, as suggested by (unreported) more general regressions. Constancy of the coefficient on this compounded variable could not be rejected, and hence it was imposed. In accordance with Honohan and Flynn (1986), an (unreported) dummy was included to remove a large outlier in 1981:4.

19

Such a regime shift might of course affect the quality of the instruments (the degree of correlation with the endogenous variable) while leaving their validity intact. Some allowance for this possibility was made by allowing shifts of coefficients both in the estimated equations (7)-(8) and in the corresponding auxiliary regressions in 1979, when Ireland joined the EMS.

20

Tests of the second type were also conducted adding the lagged values as instruments rather than as exogenous variables.

21

Pointing to a different conclusion, Grilli (1989) reports a positive and significant correlation between inflation and the revenue/GNP ratio in Ireland. Roubini and Sachs (1989), however, find that the correlation is negative and insignificant when a time trend is included, and Dornbusch (1989b) presents evidence suggesting that Ireland’s inflation rate is uncorrected with the level of its (perhaps more relevant) marginal income tax rate.

22

The real effective appreciation of 1985-86 reflected the decline of the U.S. dollar rather than a deliberate Irish exchange rate policy. This loss of competitiveness was recouped entirely in subsequent years, helped by the 8 percent devaluation of the Irish pound within the ERM in August 1986. In the first quarter of 1989, the real effective exchange rate based on consumer prices stood 5 percent below the peak reached in the first quarter of 1983; developments in competitiveness based on labor costs were substantially more advantageous (see International Monetary Fund (various years) and Mayer (1989)).

23

There are four independent post-1979 coefficients and one ovsridentifying instrumental variable; starting the estimation with a sample up to 1980:4 thus leaves three degrees of freedom to begin with. For this exercise the full-sample coefficient estimate was imposed on the dummy for 1981:4.

24

A Bayesian counterpart to this classical approach is applied by Baxter (1985) to the credibility of exchange rate policy in Chile and Argentina during the 1970s.

25

As noted by Dornbusch (1989a), in addition to forgoing the opportunity to devalue the Irish pound at both ERM realignments during 1982, the Government in that year also took steps intended to set in motion a process of fiscal adjustment. But the contribution of fiscal policy to the disinflation (and its credibility—see Section III) should not be exaggerated; in fact, the public debt/GNP ratio increased rapidly and continuously from the start of the EMS until 1988, except for 1982, when the ratio remained constant.

26

This increase also reflected the fall in government revenue from money seigniorage that resulted from the successful disinflation.

27

Persson and van Wijnbergen (1988) find, at a theoretical level, that wage controls may help to establish credibility for a disinflation program, but only if accompanied by recessionary monetary policy. They assume that wage controls carry microeconomic costs, but they do not consider the case—apparently relevant in the Irish context—in which willingness on the part of the social partners to accept wage controls is itself related to the credibility of the disinflation.

28

Suggestive empirical support for the proposition that the ERM has allowed inflation-prone participants to recoup temporary competitive losses can be found in Mayer (1989).

  • Collapse
  • Expand
IMF Staff papers, Volume 37 No. 1
Author:
International Monetary Fund. Research Dept.