II. Aspects of the Monetary Transmission Mechanism17
A. Introduction and Overview
45. Portugal’s macroeconomic performance during the past decade and a half has been marked by a successful process of nominal and real convergence. Inflation, which stood at more than 20 percent in the early 1980s, has been reduced to just above 2 percent. The success of the disinflation effort can be better gauged by comparing the Portuguese performance against the rest of the EU: the inflation differential between Portugal and the EU average has fallen from a peak of some 22 percentage points in 1994 to virtually zero at present. At the same time, economic growth in Portugal, while exhibiting higher volatility relative to most other European countries, has on average been quite strong, and served to raise GDP per capita from half of the EU average prior to EU accession to some 70 percent at present.
46. The contribution of monetary policy to this performance has been the subject of some public debate, with a broad consensus on a firm, stability-oriented conduct of policy emerging only more recently, as the benefits of such a conduct have become evident. In this context, it would appear worthwhile to attempt to obtain a quantitative sense of the impact of monetary policy on the Portuguese economy. This chapter utilizes an unrestricted vector autoregression (VAR) methodology to characterize some of the patterns of the monetary transmission mechanism in Portugal. The objective is to examine how the monetary authorities’ instruments affect some key variables (nominal and real) that may be viewed as the ultimate targets of economic policy, as well as some of the channels through which these effects operate (notably the exchange rate channel).
47. The conduct of monetary policy in Portugal underwent a fundamental shift in the late 1980s, when, under the experience of the high inflation outcomes of the beginning of the decade, monetary and credit targeting were largely abandoned. Instead, and in line with the gradual liberalization of the financial system, the monetary authorities started to rely upon interest rates as their primary policy instrument, and the exchange rate began to play an increasingly prominent role as an intermediate target. Initially, the Portuguese central bank pursued an interest rate policy that was conducive to the support of a shadow peg of the escudo to the ECU (and the deutsche mark). Eventually, this link became a formal intermediate target with the escudo’s entry into the ERM in early 1992.
48. While Portugal shares a number of important features with other countries whose currencies participate in the ERM, one crucial difference, particularly relevant for the purposes of the present inquiry, needs to be emphasized. In the case of most of the “core” ERM currencies18 of northern Europe, the bulk of their respective economies’ international trade relations is directed toward Germany and the rest of the core ERM group. Portugal, on the other hand, while also trading heavily with the core ERM area, has at least one important trading partner, namely Spain, which is not part of this area.19 In this context, the escudo has at times been pulled in opposite directions by its ERM commitments, on the one hand, and by movements in the peseta/deutsche mark exchange rate, on the other.
49. The empirical investigation of this chapter will be conducted with two broad objectives in mind. First, from a backward-looking or historical perspective, the aim is to get a sense of the direction and magnitude of influence of monetary policy changes on prices and output, and examine some of the channels through which these effects are propagated. Second, from a more forward-looking perspective, we will try to draw some inferences, inevitably of a tentative nature, on the likely impact of EMU on the monetary transmission mechanism in the case of Portugal.
50. The remainder of the chapter is organized as follows. Section B briefly describes the methodology to be utilized. Section C discusses the empirical findings of a simple 3-variable VAR system linking domestic short-term interest rates to prices and output. Section D looks at a more complicated system, which examines separately the impact of the German and domestic premium components of the Portuguese interest rate, and also considers the importance of the effective exchange rate as a transmission channel. Section E takes a closer look at the exchange rate channel, by examining the impact of a change in German interest rates on certain key bilateral exchange rates. Section F discusses some implications of the empirical findings for the monetary transmission mechanism under EMU. Section G contains concluding remarks.
B. The VAR Methodology
51. Empirical investigation of the monetary transmission mechanism typically attempts to describe how changes in certain key variables, whether policy instruments or intermediate transmission channels, affect the nominal and real variables that constitute the policy makers’ ultimate targets. Since the application of the VAR methodology to the study of monetary transmission in the United States by Sims (1980, 1982, 1986), and its further refinement by Litterman and Weiss (1984), the methodology has become a standard empirical tool in this area, being more recently applied to a number of European countries.20
52. In essence,21 VAR models attempt to explain a set of variables in terms of the lags of all the variables under consideration. For the specific case of the monetary transmission mechanism, the list would typically include policy instruments, intermediate targets or other transmission variables, and final target variables. On the basis of the estimates obtained from the VAR system, one can compute the so-called impulse response functions, which summarize the impact of a shock to the (orthogonal component of) each variable under consideration on all the variables of the system, including itself, thus capturing the essence of the transmission of monetary impulses across the economy. The attractiveness of the VAR methodology relative to more traditional large-scale econometric models stems from its simplicity and ease of implementation (at no apparent loss of predictive power), the need to impose much fewer a priori restrictions in order to achieve identification, its rich and unrestricted lag structure which provides a good safeguard against econometric pitfalls such as spurious correlation, and the ability to provide efficient estimates on the basis of much shorter samples that do not contain policy regime switches or other significant institutional changes.
53. These advantages notwithstanding, one still needs to impose (relatively few) restrictions in order to achieve identification of a VAR model. The more standard approach, also adopted for the purposes of this chapter, entails imposing a recursive structure on the contemporaneous part of the system, while maintaining an unrestricted lag structure. While this identification scheme results in no loss in terms of predictive power, it renders inferences about the underlying economic structure rather perilous. Thus, some researchers prefer the so-called “structural” VAR approach, which imposes contemporaneous restrictions based on the implications of particular theoretical models. While this latter approach is itself subject to some pitfalls,22 it enables one to test more formally whether the data are consistent with one class of theoretical models rather than another. Accordingly, for the purposes of this chapter, inferences about competing underlying models that could describe the Portuguese economy, especially as relating to the nature of shocks that may be affecting certain key variables, will be kept to a minimum and will be made with a great deal of caution.
C. The Link Between Interest Rates, Prices, and Output: Evidence from a Simple VAR System
54. This section starts off by examining a very simple 3-variable VAR model of the monetary transmission mechanism for Portugal. This model consists of a nominal short-term interest rate as the authorities’ policy instrument, and the consumer price index and real output as target policy variables.23 Given its simple structure, the model is entirely cryptic about the channels of transmission of monetary policy impulses. The model has a 6-lag structure,24 and was estimated over the 1986-95 period using monthly data.
55. As monthly data for real GDP are not available, the output variable was constructed by quarterly real GDP interpolated on the basis of seasonally adjusted industrial production. While industrial production data are recognized to be of lower quality relative to GDP data, reliance on monthly data was on balance deemed to be preferable to the obvious alternative of using quarterly data and employing actual GDP as the output variable, as this latter alternative would have entailed moving the start date of the sample at least well into the 1970s in order to ensure adequate degrees of freedom. In turn, such a long sample would have contained periods of very different monetary policy regimes, as alluded to in the introduction—indeed, mere inspection of the extreme interest rate volatility prior to the mid-1980s would suggest that interpreting the interest rate as the monetary policy instrument would have been inappropriate. Under these conditions, estimates from the longer sample would have been highly unreliable for forecasting purposes.
56. For the purposes of identification, the contemporaneous recursive structure employed entails an ordering of the variables of the system that allows the interest rate to affect both prices and output within the month, and prices to affect only output (but not the interest rate) within the month, but restricts output to affect neither the interest rate nor prices within the month. This ordering can be justified on the basis of information lags with which economic data become available.
57. Figure 6 plots the impulse response functions of this 3-variable VAR system, with a forecast horizon of 48 months. These impulse responses describe the impact of a one standard deviation shock to each variable of the system on all other variables, including itself. The chart also plots one standard deviation confidence intervals around the impulse responses.25
Figure 6.Portugal: Impulse Response: Simple VAR
Definition and source of variables can be found in the Data Appendix.
58. A number of features of the impulse response functions of Figure 6 deserve comment. It should first be noted that a positive interest rate shock exerts a dampening effect on output. The effect turns statistically significant about a year after the shock and is quite persistent, lasting throughout the forecast horizon. Other features of Figure 6 are also suggestive. Thus, shocks to output during the sample period considered appear to have been predominantly of a demand side nature. Although the specification employed does not allow one to distinguish between real and nominal demand shocks, a positive shock to output has tended to be associated by a significant rise in prices and a tightening of interest rates by the monetary authorities. This, however, should not be interpreted as implying that supply shocks have been absent during the period under consideration. In fact, as regards shocks to prices, supply side shocks appear to dominate historically: thus, a positive shock to prices appears to entail, following a brief period after the shock, a dampening (albeit only marginally significant) effect on output and an eventual loosening of interest rates by the monetary authorities. Given that relevant structural identifying restrictions26 have not been imposed in the specification, one cannot say much more about the relative magnitude of supply versus demand shocks.
59. A somewhat puzzling feature of Figure 6 relates to the estimated impact of an interest rate shock on prices. Thus, a positive shock to the Portuguese short-term interest rate is estimated to initially entail an increase in the CPI above its baseline path, with the effect being statistically significant for over a year. While prices are estimated to start falling eventually, and the point estimate falls below the baseline path two years after the shock, this latter effect never becomes statistically significant. This behavior of prices in response to an interest rate increase could thus raise doubts about the traditional view of an interest rate hike as an anti-inflationary policy action.
60. While the above behavior of prices in response to an interest rate change may appear counter-intuitive, it is by no means uncommon in the empirical literature on the transmission mechanism—it is sometimes referred to as the “price puzzle.” Indeed, this observed counterintuitive effect in the case of Portugal is much less extreme compared to the findings for other EU countries, where the price level has been estimated to remain above its baseline path over much longer horizons in response to a positive interest rate shock.27 A typical explanation, consistent with the interpretation of an interest rate innovation as a monetary policy shock, rests on the assumption of superior information about inflationary pressures on the part of the monetary authorities relative to other economic agents (and also relative to the information content of the variables of the system under consideration). In such circumstances, in a situation where the monetary authorities know that inflationary pressures are likely to emerge in the future, they may engage in a pre-emptive interest rate tightening that would tend to be associated with an increase in prices above the VAR baseline (albeit by less than if the policy, action had not been taken) and by a contraction in output.
61. It should be pointed out that the effect described above should not be interpreted as implying that positive shocks to the real interest rate did not have an anti-inflationary impact in the case of Portugal during the period under consideration. In fact, given the very sharp decline in inflation during this period (at the start of the sample inflation was still as high as 12 percent), movements in the nominal short-term interest rate may have been dominated by movements in inflation rather than by movements in the real rate, in ways not fully captured by the lag structure employed. In that sense, the nominal short-term interest rate, which is typically utilized in the literature on the transmission mechanism as the monetary policy instrument, may be an imperfect indicator of the stance of monetary policy. In addition, the simple structure of the system considered in this section, and in particular the noninclusion of the relevant transmission channels (especially the exchange rate), may also impair the interpretation of an interest rate shock. Specifically, the framework of this section would not adequately capture contemporaneous causality chains running from exchange market pressure (partly linked to inflationary shocks), to escudo depreciation and interest rate hikes to limit this depreciation, and finally (possibly with a lag) to prices and output.
62. In view of these considerations, the 3-variable VAR system was re-estimated, substituting the real for the nominal short-term interest rate. The corresponding impulse response functions are presented in Figure 7. The estimation results summarized in Figure 7 provide a very different picture of the impact of the interest rate variable on prices relative to those of Figure 6. In effect, the “price puzzle” disappears: a positive shock to the real interest rate is now associated with a steady decline of the CPI below its baseline path, with the effect remaining statistically significant throughout the forecast period.
Figure 7.Portugal: Impulse Response: Simple VAR with Real Interest Rate
Definition and source of variables can be found in the Data Appendix.
63. Most of the remaining conclusions of Figure 6 remain valid under the new specification as well. Thus, a positive shock to the real interest rate is once again estimated to exert a depressing effect on output, which turns statistically significant a year after the shock. In addition, shocks to output again appear to be predominantly on the demand side, with a positive output shock associated with an increase in prices, and shocks to prices predominantly on the supply side, with output moving in the opposite direction.
D. An Extended VAR System
64. While the estimation results of the 3-variable VAR system of the previous section provide a broad picture of the patterns of monetary transmission during the past decade, the simplicity of the specification inevitably leaves some important questions unanswered. In particular, the model of the previous section is silent about the channels through which monetary policy impulses are transmitted to the economy. In addition, it fails to do adequate justice to the specificities of the Portuguese monetary policy “regime” during the period under consideration. This section attempts to address these shortcomings, at least partially, by extending the underlying framework in two directions: a decomposition of the short-term interest rate variable into an anchor currency and a domestic premium component, and explicit consideration of the exchange rate as a transmission channel.
65. The first extension of the VAR system of the previous section relates to the specification of the interest rate variable. Instead of utilizing the overall Portuguese short-term interest rate, we consider separately the impact of two components of this rate: the anchor currency interest rate, captured by the German short-term rate, and the domestic premium over the German rate. This specification would appear to capture better the workings of the monetary policy regime during the period under consideration, that is, the initial informal pegging of the escudo and its subsequent formal participation in the ERM. In this setting, maintenance of the escudo’s peg would suggest that the monetary authorities should follow closely German interest rate changes, while relying on changes in the premium for the defense of the peg during periods of exchange market pressures. As such, it can be expected that the two components of the Portuguese interest rate would be driven by two very different sets of factors, so that aggregating them for the purposes of the VAR system would entail a substantial loss of information regarding their stochastic properties.
66. In addition, a theoretical argument could be made that the impact of the two components of the interest rate on some of the intermediate transmission variables and final target variables under consideration may be very different, so that their separate consideration would be warranted.28 First, changes in the two components of the short-term interest rate could have quite different effects on the yield curve. Thus, an increase in the German short-term rate can be expected to have a strong impact on German (and hence Portuguese) long rates, at least to the extent that the traditional term structure effect operates. By contrast, an increase in the premium to provide short-term support to the escudo may have less of an effect on Portuguese long-term rates, to the extent that longer-term credibility of the exchange rate peg is maintained. Even if the exchange rate pressure reflects market doubts about the longer-term sustainability of the peg, so that Portuguese long-term rates do go up, it would be reasonable to assume that the escudo’s expected depreciation would be associated with increased inflationary expectations, so that the ex ante real interest rate would rise by less than the nominal long-term interest rate.29
67. Second, and perhaps more important for the case at hand, the impact of the two components of the short-term interest rate on the escudo’s exchange rate could be quite different. On the one hand, an increase in the premium could entail an effective appreciation of the escudo, to the extent that Portugal has historically been more successful in supporting its deutsche mark exchange rate during episodes of ERM-wide turmoil relative to other noncore countries that constitute important trading partners. On the other hand, an increase in German interest rates could have less of an effect on the escudo’s effective exchange rate, to the extent that other ERM participants have been prepared to follow with an interest rate tightening of their own in response.
68. Third, it could be argued that, given the persistence, well into the 1990s, of substantial inflation differentials relative to the core ERM economies, the task of disinflation imposed a heavier burden on monetary policy in Portugal compared to this latter group of countries. In this context, and notwithstanding the increased importance of the exchange rate as an intermediate target, the Portuguese monetary authorities may have been at times reluctant to immediately follow an easing of interest rates in Germany. At the same time, with the persistence of an appreciable inflation differential putting pressure on the escudo’s deutsche mark peg during episodes of ERM-wide strain, the Portuguese monetary authorities could resort to increases in the premium, both to limit the extent of the depreciation and to dampen its inflationary impact. Under these conditions, changes in the interest rate premium may have at times provided a stronger signal of the stance of monetary policy relative to the anchor currency interest rate, with the bulk of the effects in question propagated directly from changes in the premium to prices and output.
69. The second extension of the framework of the preceding section aims at providing some insight into the channels through which monetary policy shocks have been propagated during the period under consideration. While the shortness of the sample limited the potential number of such channels that could be examined,30 the exchange rate channel clearly warranted particular attention. In the first place, it can be expected to be a strong transmission channel, given Portugal’s status as a small, open economy. Second, the nature of the monetary policy regime, which aims at “exchange rate stability” within the ERM as the intermediate target to achieve price stability, indicates that exchange rate developments have an important influence on monetary policy actions. Finally, it would be interesting to get a sense of the Portuguese economy’s sensitivity to autonomous (that is, not related to interest rates) exchange rate shocks.
70. That said, the question remains as to which exchange rate to use for the purposes of this investigation. Unfortunately, there is no unambiguous answer to this question. On the one hand, given the monetary policy regime, the escudo/deutsche mark bilateral exchange rate would appear to be the relevant variable in terms of having an impact on the monetary authorities’ policy instrument. On the other hand, it could be argued that the escudo’s effective exchange rate would be more relevant as a channel of transmission of monetary policy shocks to the target policy variables, that is, prices and output. In view of these considerations, two alternative specifications were estimated, one involving the escudo’s nominal effective exchange rate and the other the escudo/deutsche mark bilateral rate as the exchange rate transmission variable. Due to space considerations, only the results of the former specification are presented and discussed in this section; the impulse response functions based on the latter specification are presented in Figure 9.31
Figure 8.Portugal: Impulse Response: VAR: Extended System
Definition and source of variables can be found in the Data Appendix.
Figure 9.Portugal: Impulse Response: VAR: Extended System with Escudo/Deutsche Mark Rate
Definition and source of variables can be found in the Data Appendix.
71. The new 5-variable system was also estimated using monthly data over the period 1986-95. Figure 8 reports the impulse response functions associated with the model, showing the impact of a one standard deviation shock to each variable to all other variables of the system (including itself) over a period of 48 months.
72. The impulse response functions of Figure 8 contain some noteworthy features. In the first place, they shed some light on the policy reactions of the monetary authorities in response to a monetary policy action in Germany. Within the framework of the VAR under consideration, a shock to the German short-term interest rate is estimated to be quite persistent, with the interest rate remaining (statistically) significantly above its baseline path for almost 30 months after the shock. In this context, the Portuguese monetary authorities can be seen to react quickly, with the premium returning to its baseline path within two months after the German interest rate shock, and remaining at its baseline level throughout the simulation period.
73. The estimation results of Figure 8 suggest that the decomposition of the Portuguese short-term interest rate into its anchor currency and domestic premium components is indeed quite fruitful, as shocks to each of the two components are estimated to have quite different effects on prices and output. Thus, an increase in the interest rate premium is estimated to lead to a substantial fall in both prices and output below their respective baseline path; these effects remain statistically significant throughout the simulation horizon. By contrast, a one standard deviation positive shock to the German interest rate is estimated to entail virtually no change in output throughout the simulation horizon, while its impact on prices is once again counterintuitive, with the CPI remaining significantly above its baseline path for some 30 months after the German interest rate shock.
74. The strong estimated impact of changes in the interest rate premium on output would appear to provide some insight into the variability of output growth in Portugal, which—as noted in Chapter I of this paper—has been found to be higher than other EU countries. Thus, frequent recourse to the interest rate instrument to provide support to the exchange rate, and the subsequent reductions in the interest rate premium as exchange market pressures subsided, may have been an important factor accounting for the observed volatility in output growth. While it could be argued in this connection that a small, open economy like Portugal is inherently more sensitive to industry-specific shocks that could account for higher output variability, it should be noted that some relatively small core ERM countries, such as Belgium and the Netherlands, have experienced much more stable growth rates during the 1990s—compared to Portugal as well as to the larger core ERM economies.
75. Given the very different estimated response of output to changes in the German interest rate and the domestic premium, it is of interest to explore which factors could account for this asymmetry. In this regard, the estimation results of Figure 8 would suggest that the different impact of these two types of interest rate changes on the escudo’s effective exchange rate could be an important part of the story. Specifically, changes in the interest rate premium were found to have a much more persistent impact on the effective exchange rate compared to corresponding changes in the German component of Portuguese short-term interest rates. Thus, a one standard deviation positive shock to the interest rate premium was estimated to entail a strong effective appreciation, which remains statistically significant some 36 months following the shock. By contrast, a one standard deviation positive shock to the German short-term interest rate was estimated to result in a much less persistent effective appreciation, with the effect turning statistically insignificant less than a year following the shock.
76. That said, the estimated effective appreciation of the escudo, albeit rather small and transitory, in response to an increase in the German interest rate deserves some comment. This result is particularly interesting given the finding reported in Figure 9 (in which the bilateral escudo/deutsche mark rate is used as the exchange rate transmission variable) that the escudo tends to depreciate vis-à-vis the deutsche mark in the wake of an increase in the German short-term interest rate. This in turn raises the question as to which currencies the escudo appreciates against, in order to more than offset its depreciation against the German currency. This question, which provides some insight both into past patterns of monetary transmission, highlighting in a sense Portugal’s “positioning” in-between the core ERM countries and Spain as far as exchange rate movements go, and into the implications of EMU for Portugal, is taken up in the following sections.
77. While the estimation results suggest a very different response of output to changes in the two components of the short-term interest rate, it should be acknowledged that the interpretation of a change in the anchor currency interest rate component is not unambiguous. Specifically, the VAR system of this section cannot claim to offer a satisfactory account of the determinants of German short-term rates. To the extent that such omitted variables also have an impact on some variables of the system under consideration, interpreting a shock to the German rate as a “true” innovation in German monetary policy would appear somewhat problematic. For instance, if strong economic activity in the core ERM area is prompting German monetary policy tightening while at the same time boosting exports in Portugal, one could observe higher German rates to be associated with higher prices and stronger activity in Portugal, but this association should not be interpreted as carrying causality indications. Indeed, estimation of a system that controls for economic activity in the core ERM32 revealed that a shock to the German interest rate entails a small (but statistically insignificant) dampening impact on Portuguese output, while also going some way toward addressing the “price puzzle” mentioned above.
78. The impulse response functions of Figure 8 also provide some insight into the impact of “autonomous” (i.e. orthogonal to interest rates) shocks to the escudo’s effective exchange rate. A one standard deviation appreciation of this exchange rate is estimated to have a strong and persistent depressing effect on both prices and output. The effect on prices is felt on impact, and the CPI remains significantly below its baseline path throughout the simulation horizon. The effect on output turns statistically significant some 6 months after the shock, and similarly remains statistically significant almost throughout the remainder of the forecast horizon. The strength of these effects highlights the status of Portugal as a small, open economy.
79. Finally, the estimation results reported in Figure 8 tend to modify somewhat the conclusions regarding the relative importance of demand versus supply shocks reached in the previous section. Thus, the estimated impact of a positive shock to output in Portugal once again points to the dominance of demand-side shocks, as it is associated with a persistent (and statistically significant) increase in the price level above its baseline path. What appears somewhat problematic with regard to this interpretation is the behavior of the exchange rate, as the positive shock to output is estimated to entail an effective depreciation of the escudo for about a year after the shock. The “demand-side” interpretation may, however, be salvaged to the extent that demand shocks have originated predominantly in the area of fiscal policy during the period under consideration: empirical results from a number of EU countries suggest that episodes of fiscal expansion have often tended to undermine exchange rate credibility, leading to a weakening of the currencies concerned. On the other hand, the results of the 5-variable VAR of this section fail to bear out the conclusion of the previous section that shocks to the price level seem to reflect predominantly supply-side shocks. The impact of a one standard deviation shock to the price level on output is very weak and turns out to be statistically insignificant almost throughout the forecast horizon.
E. A Closer Look at the Exchange Rate Channel
80. The empirical results of the previous section raised some questions about the impact of a shock to the German interest rate, as transmitted by the exchange rate channel. Specifically, a positive shock to the German rate, while estimated to entail a depreciation of the escudo vis-à-vis the deutsche mark, was also estimated to result in the Portuguese currency’s effective appreciation. This would suggest, by definition, that the escudo must be appreciating against some other currencies, to such an extent as to more than offset its depreciation against the deutsche mark.
81. A brief investigation of the direction of impact of interest rate changes on a number of bilateral escudo exchange rates would appear worthwhile for a number of reasons. From a backward-looking perspective, it could provide a sense of the implications for Portugal of the combination of ERM participation and close trade integration with a noncore ERM country. From a more forward-looking perspective, it may provide indications about monetary transmission under EMU, when certain key bilateral exchange rates will be permanently locked—an issue which will be taken up in the next section. From both of these perspectives, it would be interesting to determine whether the bulk of the escudo’s effective appreciation in the wake of a positive shock to the German short-term interest rate is related to an appreciation of the deutsche mark bloc against non-ERM currencies, or whether it occurs against noncore ERM currencies.
82. To gain some insight into these questions, a 4-variable VAR system is specified and estimated. The system includes the following variables: the German short-term interest rate; the deutsche mark/dollar exchange rate, as a proxy for exchange rate movements of the deutsche mark bloc against non-ERM currencies; the escudo/deutsche mark exchange rate; and the escudo/peseta exchange rate, to capture movements in the escudo vis-à-vis a noncore currency of a key trading partner. The impulse response functions associated with this VAR system are reported in Figure 10.
Figure 10.Portugal: Impulse Response: VAR: Bilateral Exchange Rates
Definition and source of variables can be found in the Data Appendix.
83. The empirical results of Figure 10 reveal certain interesting patterns regarding the impact of changes in the German short-term interest rate. In the first place, they confirm the earlier result that a one standard deviation positive shock to the German short-term rate entails a depreciation of the escudo against the deutsche mark. This effect remains statistically significant throughout the forecasting horizon. Second, the same German interest rate shock results in an appreciation of the deutsche mark vis-à-vis the dollar. While this effect turns marginally significant after some 6 months, its magnitude is quite small. In fact, combined with the escudo’s depreciation against the deutsche mark, this suggests that a “deutsche mark bloc effect” cannot account for the escudo’s estimated effective appreciation.
84. Rather, the dominant source of this effective appreciation is revealed by the estimated impact of the shock to the German rate on the escudo/peseta bilateral rate. Thus, following the shock, the escudo is estimated to appreciate against the Spanish currency. The effect occurs virtually on impact, is quite large, and persists throughout the forecasting horizon. It would thus appear that it is this last effect which constitutes the source of the escudo’s effective appreciation, despite its depreciation against the deutsche mark.
85. The effects just discussed provide an important insight into the patterns through which monetary policy impulses have been transmitted under the ERM in the case of Portugal during the period under consideration. The empirical results obtained above suggest that a monetary contraction in the anchor country has tended to exacerbate exchange rate pressures on the currencies of the noncore countries. In the case of Portugal, such a monetary shock has tended to create a rather tricky situation, given that a large noncore ERM country, namely Spain, is a key trading partner. The impulse response functions of Figure 10 indicate that, on average, the escudo has been steered somewhere in-between the core currencies and the peseta in the wake of such policy shocks. In fact, it would appear that, under these conditions, the escudo has generally appreciated against the peseta by more than it depreciated against the deutsche mark, thus accounting (given the relative importance of Germany and Spain as Portugal’s trading partners) for the currency’s small effective appreciation, as documented in the empirical results of the previous section.
86. The estimated impact of an autonomous shock to the deutsche mark/dollar rate, on the basis of Figure 10, is also worth noting. Research on a number of European countries whose currencies participate in the ERM has tended to indicate that, at least in certain cases, changes in the deutsche mark/dollar cross rate has tended to affect the bilateral rates of these currencies vis-à-vis the deutsche mark.33 Specifically, an appreciation of the deutsche mark against the dollar has sometimes been found to be also associated with a weakening of a number of other ERM currencies against the deutsche mark, an effect possibly reflecting the reserve currency status of the dollar and the deutsche mark. The empirical results of this section suggest that such an effect may also have been relevant for Portugal, at least to some extent, during the period under consideration. Thus, the impulse response functions of Figure 9 indicate that a positive shock (appreciation) of the deutsche mark against the dollar has tended to entail a depreciation of the escudo against the German currency. While the effect is estimated to occur on impact, it appears to be very short-lived, turning statistically insignificant some 4 months after the shock.
F. The Implications for EMU
87. A question that naturally arises relates to whether the empirical results obtained so far can shed some light on the likely impact of EMU on the monetary transmission mechanism in the case of Portugal. It should be emphasized at the outset that this question needs to be approached with a great deal of caution. After all, EMU would constitute a fundamental policy regime switch, entailing a new institution charged with the conduct of monetary policy under potentially different policy objectives and utilizing potentially different policy instruments. In this context, the behavior of economic agents, which presumably has been reflected in the empirical results of this chapter, could change in fundamental ways. In that sense, relying on empirical results obtained on the basis of a historical sample to make inferences about the monetary transmission under EMU could be quite hazardous.34 Furthermore, even if behavior were to remain fundamentally unchanged, EMU could entail major changes in the distribution of shocks affecting the economy. For instance, some recent research has tended to indicate that an important share of (asymmetric) demand shocks in the ERM area can be traced to fiscal policy. It would appear particularly difficult to predict the size, and even the direction, of changes in the distribution of such shocks under the post-EMU Stability and Growth Pact.
88. With these important caveats in mind, this section addresses the potential impact of EMU on the transmission mechanism from a rather narrow perspective. Specifically, it utilizes the empirical results obtained to inquire about the likely impact of the elimination of the interest rate premium (or at least that portion of the premium that reflects expectations of exchange rate changes) and the permanent locking of the cross exchange rates between prospective EMU members. On both accounts, it could be postulated that the impact of EMU in the case of Portugal may be rather benign.
89. The empirical results of this chapter pointed to a rather strong impact of changes in the premium over the anchor currency interest rate (compared to changes in the anchor currency interest rate itself) on economic activity in Portugal. On this basis, one could postulate a link between the high volatility of the interest rate premium (which has constituted the main instrument to provide support to the exchange rate) and the variability of output growth—which has tended to be higher in the case of Portugal compared to most other EU countries. Under these conditions, it could be argued that the virtual elimination of the interest rate premium under EMU could entail a more stable pattern of output growth. In this regard, however, the estimation results of Figure 8 should be interpreted with particular care, as the impact of the euro interest rate under EMU could be much stronger than the corresponding impact of the anchor currency interest rate under the ERM. Indeed, one could reasonably postulate that, with currency unification, the euro interest rate would capture the bulk of the signaling role previously embodied in the interest rate premium. That said, one could still expect on balance lower output variability on account of the elimination of the premium, on the basis of the much higher volatility of the premium component relative to the anchor currency interest rate component of Portuguese short-term rates under the ERM.35 The likely magnitude of this effect would of course depend on the quantitative importance of the interest rate premium relative to other (that is, not related to interest rates) sources of economic fluctuations—a question that was not explored in this chapter.36
90. Perhaps the most often cited potential cost of EMU relates to the loss of the exchange rate instrument to dampen the impact of asymmetric shocks. At first glance this argument may appear applicable to Portugal, given that the escudo has undergone substantial fluctuations in the past, at least in comparison to the core-ERM currencies. The empirical results of this chapter could, however, be interpreted as suggesting that the potential costs involved may be smaller than is often thought, given the way that Portuguese monetary policy has in practice operated in the past.
91. To illustrate this point, one could consider an inflationary shock elsewhere in the euro area that prompts the European Central Bank to tighten interest rates. In such circumstances, under EMU, Portugal would no longer have the option of exchange rate depreciation to offset a potentially adverse impact on economic activity. By contrast, under the ERM arrangement, it could resort to depreciation against the deutsche mark following an increase in German interest rates. It should not be overlooked, however, that under these same circumstances, the escudo has tended to appreciate against certain key non-core currencies. Indeed, the empirical results of this chapter suggest that the escudo’s appreciation against non-core currencies following a tightening of German monetary policy has tended to offset (in fact, more than offset) its depreciation against core currencies. In a sense, the loss of Portugal’s option to depreciate against the deutsche mark should be balanced against the loss of this same option on the part of other noncore countries.37 Given the relative degrees with which the depreciation option had been exercised in the past by the various countries concerned, as have been estimated in this chapter, the actual cost of the loss of the exchange rate instrument in the case of Portugal may thus in fact be rather limited.
7. Concluding Remarks
92. This chapter explored some of the patterns of the monetary transmission mechanism in Portugal. The main conclusions can be summarized as follows:
• Monetary policy actions, as captured by shocks to the short-term interest rate, turn out to have strong effects on key macroeconomic variables, notably prices and economic activity. The exchange rate was found to constitute an important transmission channel regarding these effects—a reflection of the openness of the Portuguese economy.
• Additional insight into the nature of these effects can be gained by decomposing the Portuguese short-term interest rate into an anchor currency and a domestic premium component. Changes in the latter component are found to entail a much stronger impact on output and prices compared to changes in the former. This stronger impact of the premium in turn appears, at least partly, to reflect a more persistent, effective appreciation of the escudo in the wake of a positive premium shock.
• Nonetheless, a positive shock to the German short-term interest rate is also found to entail some effective appreciation. This occurs despite the depreciation of the escudo vis-à-vis the deutsche mark in the wake of the tightening of German monetary policy, as this effect has been typically (more than) offset by the escudo’s appreciation against noncore ERM currencies—an important aspect of Portuguese monetary and exchange rate policy since the late 1980s.
93. On the basis of the empirical results, certain, highly tentative, inferences may be drawn regarding the impact of EMU on the monetary transmission mechanism. In the first place, elimination of the interest rate premium under EMU can be expected to reduce output volatility. Second, the historically very different behavior of the escudo relative to other non-core currencies in response to a tightening in the anchor currency interest rate mentioned above would suggest that the loss of the exchange rate instrument in the face of asymmetric shocks may entail a smaller cost in the case of Portugal than may appear at first sight.
94. The above considerations should not, however, mask the considerable uncertainties which remain with regard to the transmission mechanism under EMU. This in turn would suggest that caution needs to be applied with regard to conclusions on the impact of EMU on the economy’s sensitivity to interest rate changes. Indeed, it is quite conceivable that closer economic integration of Portugal in the euro area, and the entrenchment of low and stable inflationary expectations, could render the economy more sensitive to changes in the euro interest rate than it has been with regard to changes in the deutsche mark interest rate under the ERM arrangement.
All variables, except interest rates, are in natural logarithms. All series were drawn from the International Financial Statistics (IFS) published by the Fund, except for data on quarterly GDP, which were drawn from the OECD’s Analytical Database.
|PRSH:||Portuguese call-money interest rate.|
|GRSH:||German call-money interest rate.|
|NEER:||Nominal effective exchange rate for Portugal.|
|CPI:||Portuguese consumer price index.|
|NGDPM:||Portuguese GDP (at constant prices) interpolated using monthly industrial production figures.|
|LNDMESC:||Deutsche mark/escudo exchange rate (deutsche mark per escudo).|
|LNDMUS:||U.S. dollar/deutsche mark exchange rate (US dollars per deutsche mark).|
|LNPSESC:||Peseta/escudo exchange rate (pesetas per escudo).|
|RPSH:||Real Portuguese call-money interest rate, deflated by CPI inflation.|
BarranF.V.Coudert and B.Mojon1996“The Transmission of Monetary Policy in the European Countries” CEPII Document de Travail No. 96-03February.
BlanchardO. J. and D.Quah1989“The Dynamic Effects of Aggregate Demand and Supply Disturbances,”American Economic ReviewVol. 79No. 4.
DaleS. and A. G.Haldane1993“A Simple Model of Money, Credit, and Aggregate Demand,”Working Paper No. 7Bank of England.
DiasF.C.J. A. F.Machado and M. R.Pinheiro1996“Structural VAR Estimation with Exogeneity Restrictions”Oxford Bulletin of Economics and Statistics 58 2.
EscriváJ. L. and A. G.Haldane1994“The Interest Rate Transmission Mechanism: Sectoral Estimates for Spain,”Working Paper No. 9414Bank of Spain.
FaustJ. and E. M.Leeper1994“When Do Long-Run Identifying Restrictions Give Reliable Results?,”Board of Governors of the Federal Reserve System International Finance Discussion PapersNo. 462March.
GarretsenH. and J.Swank1994“The Transmission of Interest Rate Changes and the Role of Bank Balance Sheets: A VAR Analysis for the Netherlands,”The Netherlands Bank MEB SeriesNo. 4.
GiavazziF. and A.Giovannini1986“The EMS and the Dollar”Economic PolicyApril.
HalikiasI. and J.Levy1996“Aspects of Monetary Policy”in IMF Staff Country ReportNo. 96/80August (Washington: International Monetary Fund).
LevyJ. and I.Halikias1997“Aspects of the Monetary Transmission Mechanism under Exchange Rate Targeting: The Case of France,”IMF Working Paper WP/97/44April (Washington: International Monetary Fund).
LittermanR. and L.Weiss1985“Money Real Interest Rates and Output: A Reinterpretation of Postwar U.S. Data”Econometrica 53.
LucasR. E. Jr.1976“Econometric Policy Evaluation: A Critique” in K.Brunner and A. H.Meltzer“The Phillips Curve and Labor Markets”New York: North Holland.
SimsC. A.1980“Macroeconomics and Reality”Econometrica 48.
SimsC. A.1982“Policy Analysis with Econometric Models”Brookings Papers on Economic Activity.
SimsC. A.1986“Are Forecasting Models Useful for Policy Analysis?”Federal Reserve Bank of MinneapolisQuarterly ReviewWinter.
Prepared by Ioannis Halikias and Joaquim Levy.
It should be stressed that the distinction between “core” and “noncore” currencies made in this chapter has no normative implications, that is, it does not refer to possible differences in policy objectives. Rather, it is a positive distinction relating to the actual historical degree of exchange rate fluctuation around a currency’s central parity.
In this regard, the Portuguese situation is not unlike that of Ireland.
See, for instance, Dale and Haldane (1993) on the United Kingdom; Escrivá and Haldane (1994) on Spain; Garretsen and Swank (1994) and Halikias and Levy (1996) on the Netherlands; Levy and Halikias (1997) on France; as well as Barran, Coudert and Mojon (1996) on selected European countries (unfortunately excluding Portugal).
See, for example, Faust and Leeper (1994) for a critique of “structural” long-run identification restrictions.
In this chapter, all variables are expressed in natural logarithms, except for interest rates which are expressed in percent. Variable definitions, notation, and data sources are provided in the Appendix.
The length of the lag structure was chosen on the basis of the Akaike criterion.
The confidence intervals were computed on the basis of Monte Carlo draws.
These restrictions would have to be long-run restrictions of the type first considered by Blanchard and Quah (1989).
In fact, if purchasing power parity holds, there would be no effect on the real long-term rate.
Transmission channels typically considered in the literature include long-term interest rates, as well as money and credit aggregates.
Both the effective and bilateral rates of the escudo have been defined so that an increase stands for an appreciation and a decrease for a depreciation.
This was done by including German GDP growth alternatively as an additional endogenous variable or as an exogenous variable in the VAR system—because of the recursive contemporaneous structure of the model, the statistical issues arising from exogeneity discussed in Dias, Machado and Pinheiro (1996) do not arise. On space considerations, the results are not reported, but are available from the authors upon request.
It should be recalled that, despite its advantages, a VAR system is still a reduced-form model, and as such remains vulnerable to the Lucas (1974) critique in the face of fundamental policy regime shifts.
The impulse responses of Figure 8 suggest that a one standard deviation shock to the premium was over six times as large as a corresponding shock to the German rate.
Chapter I of this paper postulates that economic fluctuations in Portugal (over a much longer horizon) can inter alia be attributed to shifts in the conduct of financial policies.
Of course, this argument assumes that other noncore countries will participate in EMU at exchange rates close to their current level.