When entering the Exchange Rate Mechanism of the European Monetary-System (EMS) on October 8, 1990, the United Kingdom fixed the current market rate as the new central rate within the EMS, thereby appreciating the latter by 4.9 percent. This is surprising since its economy has been afflicted by current account deficits and comparatively high inflation rates which threaten to further deteriorate its international competitiveness. Standard economic reasoning, including Keynes’ critique of England’s return to the gold standard in 1925, rather suggests that the United Kingdom, given its current account problems, should have depreciated the pound sterling when entering the Exchange Rate Mechanism of the EMS.
A similar observation, albeit in a different context, can be made when realignments within the EMS occur. Giavazzi-Pagano (1988)1/ report that, when realignments occur, excess inflation countries, e.g., Italy, “obtain devaluations which are generally insufficient to make up for the real appreciation experienced since the previous realignment” (p. 1055, see also p. 1067). Why do countries decide to maintain part of the real appreciation at realignment dates?
Another observation concerns Austria in the late 1970s. At that time, Austria was troubled with current account deficits ranging between 2-3 percent of GNP. In addition and similar to the case of the United Kingdom in 1990, higher inflation rates than those in Germany undermined Austrian competitiveness abroad. Yet, before Austria started to closely tie its currency to the deutsche mark (DM) in 1981, between 1979 and 1981 it appreciated the current market rate of the schilling by 4 percent relative to the DM.
How can this pattern of behavior be explained? An often used theory refers to the advantage of tying the central bank’s hands. The Giavazzi-Pagano (1988) paper follows this theme. As inflationary surprises by the central bank are the source of (Pareto-)inefficiency in their model, according to a Barro-Gordon type of argument (Barro-Gordon, 1983), Giavazzi-Pagano conclude that the more a central bank is constrained in creating such surprises the higher will be the welfare of the country. One such constraint is an exchange rate commitment. When such a commitment is linked with a persistent real appreciation of the currency, then the cost of creating inflationary surprises increases even more thus further reducing the incentives of the central bank to inflate the economy (pp. 1067).
There are several objections against the above reasoning. One objection (e.g., Obstfeld 1988) notes that real appreciations lead to current account deficits which, when financed via the capital account, necessitate an increase in domestic interest rates thereby making the money supply endogenous. Yet, the exogeneity of the money supply is precisely what is needed in the Giavazzi-Pagano (1988) model to make inflation possible. In addition, persistent current account deficits are difficult to sustain.
The objection of this paper is that the Giavazzi-Pagano (1988) model misses an important aspect of European monetary policy during the last decade. One can claim that Germany has successfully demonstrated to Western and Central Europe how welfare improving it is to have a “conservative” (Rogoff 1985) central bank giving a high priority to price stability. Hence, central banks in countries such as France or Italy started about a decade ago to change their policies when adopting the “conservative” stance of the German Bundesbank. This change went along with the rising distrust in Keynesian precepts. Although the new stance was indicated to the public by exchange rate commitments to the DM, the public, especially the trade unions, believed the old policies still prevailed. Real appreciations, at least initially, served to signal the new “conservative” stance.
The paper deals with the issue of signaling a new policy stance by the central bank. It will argue that it may pay off to signal the change by setting the nominal exchange rate at a level which implies an appreciated real exchange rate (e.g., below the purchasing power parity, PPP). Such an initial real appreciation demonstrates to domestic producers and unions that “conservative” preferences have been adopted by the central bank and that the exchange rate commitment, e.g., to the DM, is serious. 1/
An important point is that real appreciations are only initially needed for signaling purposes. Real appreciations need not to prevail in order to make inflation persistently costly, as in the Giavazzi-Pagano (1988) approach. The working of the signaling can be summed up by saying that the less the central bank cares about real issues and the better this message is conveyed to the domestic producers and unions the more they will do to maintain their international competitiveness. 2/3/
The theory of signaling games was first applied to the analysis of monetary policy by Vickers (1986). His analysis deals with domestic monetary policy, whereas this paper discusses exchange rate issues. In addition, because of suitable simplifications, the main points of the type of signaling games suggested by Vickers (1986) can be illustrated in an easy way. A simplified version of a signaling game will be presented in this paper to make it easier to extend the analysis.
The paper is organized in the following way: In section 2, a basic outline of the model is given, discussing the main variables and the sequencing of moves within the signaling game. Section 3 describes the two players and provides the theoretical results for a two-period game. Section 4 employs an example which demonstrates the welfare implications of the game. Section 5 extends the analysis, and section 6 concludes the paper. 1/
II. Basic Outline of the Model
The model reflects the basic macroeconomic setting used in analyzing the exchange rate problems of an open economy. Within the setting, a simplified signaling game between the central bank and the union is introduced thereby making wage formation endogenous as well as the policy itself. Fiscal policy issues, however, are left aside in order to restrict the analysis to a two-person game.
We start the analysis with six variables all denoted in logarithm: the nominal exchange rate (e), the price level of the domestically produced goods (p), the foreign price level (p*), the nominal wage rate (w), the consumer price index(q), and the real exchange rate (ρ). The two latter variables follow from the former. The consumer price index is defined as
whereas the real exchange rate ρ is defined as
Real variables such as real GNP, the level of employment, or the current account are all assumed to be uniquely related to the real exchange rate. The relationship between them is strictly monotonous and positive. For example, if the real exchange rate is appreciated (ρ < 0), real GNP, the employment level, and the current account will all decline.
To simplify the analysis, we first set w=p. This simplification is tantamount to assuming that there is only one factor of production, labor, which enters the production function in a linear, one to one way, and that domestic producers fix their prices equal to marginal cost. We assume that the trade union sets the nominal wage rate unilaterally. 1/ Secondly, for ease of exposition and without changing the general conclusions of the model, we assume that the foreign currency is inflation-free. Hence p*=0.
The time structure of the game is kept as primitive as possible. In period zero, we set the wage rate, w0, equal to 0, since only the central bank acts. It sets e0 = x, x being positive (depreciation), or negative (appreciation). By setting the exchange rate the central bank changes the consumer price index and the real exchange rate in period zero.
In period one, the trade union moves first. Before it decides on the level of the wage rate for the first period w1, it receives e0 = x as a message which it uses to assess the “type” of the central bank (“conservative” or not). When deciding upon w1 the trade union takes into account its exchange rate expectation, ee1, which in turn is based on its belief about what type the central bank is given the message e0 in period zero. We assume that the trade union has rational exchange rate expectations. After the union has set the wage rate w1 the central bank fixes the final exchange rate, e1. This move ends the game.
III. The Model
1. The trade union
We postulate a utility function for the trade union for period one, V1, from which we derive its wage equation. This utility function has two terms, which reflects the fact that the union is interested in having high real wages (w-q) and full employment. For computational convenience, the first term is assumed to be linear (the higher the real wage the higher the utility of the union), whereas the second is quadratic around the full employment-real exchange rate level (ρf) which can be the PPP-level. 2/ Note that due to the sequencing of moves in period one--the union moves before the central bank decides on the exchange rate in period one--the variables e1, p1, and thus q1 are expected by the union (written as ee1, pe1 respectively qe1):
The parameters b1 and b2 (b1 ≥ 0, b2>0) express the relative importance of the two terms. Since w=p=pe, w-q = a(w-e-p*), and since p* = 0 (and hence P*1 = 0) one can rearrange equation (3a):
By maximizing equation (3b) with respect to w1, one obtains the linear wage equation of the union:
The coefficient b summarizes the relative interest of the union in high real wages vis-a-vis full employment. For example, if the union is only interested in full employment (b1 = 0), we get b = 0. Then, any expected level of the exchange rate will be prompted by a parallel level of the nominal wage rate so that the expected real exchange rate is kept at ρf. However, if b1 increases relative to b2, the union demands higher real wages even at the price of growing unemployment.
The assumption of rational expectations requires that, in equilibrium, the union’s expected exchange rate for period one must be correct.
2. The central bank
a. Preferences and individually rational strategies
Central banks pursue two goals: a “nominal” one which consists of curbing inflation and a “real” one of promoting real growth, of maintaining full employment, etc. We assume that the goal of curbing inflation is equivalent to holding the consumer price index q0 and q1 as close as possible to zero (remember that all variables are denoted in logarithm and that zero is the starting value for the foreign and domestic price levels). The pursuit of the “real” goal is assumed to consist of keeping the real exchange rate at the full employment - real exchange rate level ρf.
For ease of calculation, a quadratic utility function will be used and it will be further assumed that the importance attached to the two goals does not change with time. Hence, we specify the utility function U by
with c≥0 as the relative weight put on the goal for attaining the full employment-real exchange rate level.
Remember that q0 = a*e0, q1 = a*e1 + (1-a)w1, ρ0 = e0, and that ρ1 = e1-w1.
Due to the additivity in equation (5) and the sequencing of moves within the model, the central bank first calculates the optimal e0 without taking into account the forthcoming wage rate level, w1. Then, in period one, it will respond to changes in the wage rate that are known to it when calculating the optimal e1. Hence, we can write for the optimal e0 and e1
Note that the parameter of w1 in equation (6b), c2, decreases monotonically, if the relative weight put on the goal for attaining the full employment-real exchange rate level, c, becomes smaller. The parameter c2 can even be negative, if a*(1-a) > c. Otherwise It will be positive. The reason for this ambiguity in sign is clear: If the central bank has a sufficiently higher interest in reducing inflation than in keeping the economy at the full employment-real exchange rate level, then it will react with an exchange rate appreciation, not with a depreciation of its currency when wages increase. If, however, c approaches infinity, then the exchange rate change (the difference between e1 and e0) reflects the wage increase (c2=1).
The first term on the right side in equations (6a) and (6b) stands for the interest of the central bank in correcting the existing real misalignment of the domestic economy. If, e.g., ρf>0, then it will depreciate the currency in period zero and maintain that amount of depreciation in period one. The amount of depreciation depends on the relative interest it has in achieving the “real” goal (if c=0, then it will do nothing to correct the real misalignment).
b. Two types of central banks
As indicated in the introduction, one can observe two types of central banks in Europe. The first is “conservative” (like the German Bundesbank) and will be called the “dry” (= truly conservative) according to an old British typology used in Vickers(1986). Its preferences are characterized by a relatively small value of the parameter c indicating its high interest in price stability. Then, as noted above, the value of c2 tends to be around 0. To simplify the analysis, we assume c2 precisely equal to 0, i.e., 1 > c = a*(1-a). The “dry” central bank while correcting an initial misalignment of the real exchange rate (to a certain degree), will do nothing when domestic wages increase. The reason for this is that its interest in maintaining full employment (which would demand a depreciation) is sufficiently outweighed by its interest in curbing inflation (which would demand an appreciation).
For a “dry” there are no costs of entering an exchange rate commitment, since--in our model--no reason for resetting the exchange rate in period one may occur. To peg the exchange rate to the DM in period zero may, hence, have the advantage of signaling the central bank’s new preferences to the union without facing the costs associated with a reneging in period one.
Contrary to the “dry” approach described above is the reaction of the second type of central bank called the “wet”. The “wet” is mainly interested in achieving the “real” goals of exchange rate policy. To easily model its situation, we assume, for the “wet”, c > 1 > a*(1-a) so that the value of both parameters, c1 and c2, is close to 1. (For the “wet”, the special parameter value of c2 will henceforth be denoted by d.)
Without exploring in detail the problem of an exchange rate commitment of the “wet” central bank, we assume that it pays off for it to pledge in period zero to keep the exchange rate constant. Thus, in period zero, the “wet” mimics the “dry”. By announcing its commitment the “wet” hopes that wage increases will not occur in period one. However, if wage increases do occur, (assuming that reneging will not be too costly for it) the “wet” central bank will renege its pledge and reset the exchange rate to partially offset the loss in international competitiveness (see equation (6b)).
Since c1 and c2 are monotonous functions of c, it may suffice to restrict the typology of central banks to the two types just stated.
3. Equilibrium with complete information
If the trade union has complete information about the type of central bank it faces, then, with rational expectations by the union (ee1 = e1), the non-cooperative equilibrium solution is (Pareto-)inefficient:
The inefficiency is like the one in Barro-Gordon (1983). It occurs in the case of the “dry” and “wet” central bank. In both cases, even if c2 = 0, the nominal exchange rate, the nominal wage rate for period one, and hence inflation, can be reduced without changing the real exchange rate.
4. Analysis of the game with incomplete information
In order to get a benchmark for the analysis we set ρf = 0. That is, we assume that there is no real misalignment before the game starts. Real misalignments may, however, occur afterwards, since due to the exchange rate and wage rate decisions the real exchange rate ρ in both periods may be unequal to 0 (=ρf). This assumption will be dropped in subsection 5.2.
As mentioned in the introduction, the main points of the type of signaling games suggested by Vickers (1986) can be depicted in simpler terms. This can be done by assuming that the strategy space of the central bank is finite and is restricted in the following way: in period zero, the “dry” can set either e0 = x < 0 (appreciate the currency to the level x) or e0 = 0. The latter response is optimal for the central bank according to equation (6a), since ρf=0. Clearly, setting e0 = x < 0 is only a second best solution. In period one, the “dry” chooses e0 = e1, since the optimal response of the “dry” in period one is independent of w1 due to c2=0. The “wet” central bank - because it mimics the “dry” in period zero, but does not stick to the exchange rate commitment in period one, when wage increases occur - sets either e0=x or e0=0 in period zero and e1=d*w1 in period one.
If desired, a third or fourth strategy could easily be added. But since x (the amount of appreciation) is variable, there is still ample generality in the analysis despite the restriction on the strategy space.
b. Analysis of the game
The analysis of the model can be summarized in four steps:
(1) The trade union has no prior knowledge about which type of central bank is sending the signal e0=0 or e0=x<0. It knows that there are four possible combinations. Two combinations consist in having both types sending the same signal, either 0 or x. These two combinations characterize a pooling situation in which an updating of the initial beliefs of the trade union (probability z for having a “wet”, 1-z for having a “dry” as central bank) is impossible since both send the same signal. Yet, although the two types are sending different signals, the union can distinguish between them and will shape its expectations accordingly.
(2) In a pooling situation (case P), in which the trade union cannot determine the type of central bank, the trade union’s expectation concerning the exchange rate is
If the two types send separate signals, the trade union will identify which central bank is “dry” (case SD) and which one is “wet” (case SW). Hence, it expects in the “dry” case (case SD)
whereas when encountering the “wet” (case SW)
with y=0 in case SD, y = z in case P, and y = 1 in case SW. To distinguish between the three cases, we write w1|y=0, w1|y=z, and w1|y=1. Remembering the restrictions on the variables and parameters in equation (6) we can rank w1: w1|y=0 < w1|y=z < w1|y=1.
(3) A (non-cooperative) equilibrium is defined by a triple (e0,e1,w1) such that (a) e0 and e1 maximize the utility function of the “dry” (UD) respectively the “wet” (UW) given w1 and (b) w1 maximizes V1 and is based on correct expectations, i.e., on rational exchange rate expectations about e1.
By requiring that the exchange rate expectation of the union be correct in equilibrium, we rule out any “cheating” by the monetary authority. Hence the equilibrium constitutes a “time-consistent” solution.
Two equilibria exist. The first one is a separating equilibrium in which it is advantageous for the “dry” central bank in period zero to separate itself from the “wet” by setting e0 = x < 0. In period one it chooses e1 = e0. The “wet” fixes e0 = 0 and e1 = d * w1. If the trade union gets as a message e0 = x, it expects ee1 = e0 = x and hence sets w1|y=0 accordingly. If the union notices e0 = 0, it expects ee1 = d * w1 and chooses w1|y=1. Notice that in this equilibrium, ee1 = e1 holds.
Although the “dry” central bank would prefer e0 = 0 given a certain wage rate, it chooses e0 = x due to the lower wage reaction when the trade union recognizes it as being “dry”.
The second equilibrium is a pooling one in which both, even the “dry”, do not appreciate the currency. Both choose e0 = 0 in period zero. In period one, the “dry” opts for e1 = e0, whereas the “wet” sets e1 = d*w1. The union sets its wage rate w1|y=z if both types choose e0=0, and w1|y=z or w1|y=1 otherwise (the latter is on the basis of off-equilibrium beliefs).
Again, in equilibrium, the wage reaction w1|y=z is based on rational expectations about e1, since the union cannot distinguish between the two types in period zero.
(4) To prove this result we start by assuming that a separating equilibrium exists. If such an equilibrium exists, it is evident that it must be the situation described above rather than the other in which the “wet” appreciates the currency (e0 = x < 0)) and the “dry” sticks to the current level (e0 = 0). The reason for this is that this latter situation would yield the most unfavorable result for the “wet” central bank: Although it sets e0 = x < 0, since expectations need to be correct in equilibrium, it is recognized by the union as “wet” (as one that will set e1 = d * w1). Consequently the wage reaction will be the highest w1|y=1. So it does not get a low wage reaction, yet, it endangers the achievement of real goals due to its choice of e0 = x < 0 instead of e0=0. Hence, for the “wet”, its utility decreases when switching to e0=x<0 which contradicts the condition for an equilibrium that e0 maximizes the utility function given a certain wage rate.
To guarantee the existence of a separating equilibrium we first need to have
This inequality sets an upper bound on |x|. Calculating this upper bound leads to a rather lengthy quadratic expression (due to the quadratic utility function U). Yet, the expression for the upper bound has the intuitively clear property that if b is strictly positive and d approaches 1 then the upper bound tends monotonously to plus infinity. If the inequality (10a) holds, then it is not optimal for the “dry” to depart from its decision to set e0 = x < 0. Secondly, the inequality
must hold, which makes it optimal for the “wet” to choose e0 = 0. This inequality sets a lower bound on |x|. Again we get a quadratic expression. Basically, b ≤ |x| must hold (sufficient condition). Note that if d is sufficiently large both inequalities,(10a) and (10b), can hold at the same time.
Essentially, the existence of a separating equilibrium presumes that the “dry’s” interest in reducing wage increases outweighs the loss incurred by appreciating the currency, whereas, for the “wet”, the cost of this initial appreciation already exceeds the benefit of getting a lower wage reaction.
If a pooling equilibrium exists, both types of central banks choose e0=0. The reason for this is that in a pooling equilibrium the trade union sets w1|y=z. Given this wage reaction, for both types of central banks it is optimal not to opt for e0 = x < 0 due to the specification of both utility functions UD and UW. Only e0 = 0 is optimal for both given w1|y=z Since w1|y=z is based on rational expectations due to the fact that both types of central banks choose the same value for e0 in period zero, a pooling equilibrium exists.
IV. An Example
In this section, the welfare implications of the different equilibria are compared. In order to avoid lengthy expressions, we choose x = -b and d sufficiently large (i.e., if a=1/3, then d>0; if a=1/2, then d>0.42). This specification suffices to demonstrate the kind of conclusions one gets from the model. If x = -b, then the following results hold:
a. for the separating equilibrium
|e0 = - b||e0 = 0|
|q0 = - (a * b)||q0 = 0|
|ρ0 = -b||ρ0 = 0|
|w1 = 0||w1 = b / (1-d)|
|e1 =e0 = -b||e1 = d * w1|
|q1 = q0||q1 = w1* (a*d + 1-d)|
|ρ1 = -b||ρ1 = (d-1) * w1 = -b|
b. for the pooling equilibrium
|e0 = 0||e0 = 0|
|q0 = 0||q0 = 0|
|ρ0 = 0||ρ0 = 0|
|w1 = b /(1 - 0.5 d) > b||w1 = b /(1 - 0.5 d) > b|
|e1 = e0||e1 = d * w1|
|q1 = (1- a) * w1||q1 = w1* (a*d + 1-d)|
|ρ1 = - w1 < - b||ρ1 = (d - 1) * w1 < 0|
Several observations can be made when comparing the different results:
1. A comparison of the “dry” versus the “wet” approach in the separating equilibrium: In period zero, the “dry”, but not the “wet” appreciates the currency in nominal as well as in real terms. Yet in period one (which can be interpreted as the long run), ρ1, the real appreciation is the same for both types of central bank.
This result is remarkable insofar as, In the long run, the unemployment levels are the same, although the “dry” manages to have a better record concerning inflation (lower w1 and q1). Therefore, the “dry” has policy results In the long run which strictly dominate those of the “wet”.
2. The “dry” approach in a separating and a pooling equilibrium: Clearly, in period one, the “dry” is better off in the separating than in the pooling equilibrium. In the latter equilibrium, the inflation as well as the real appreciation is higher than in the former. Consequently, for the “dry”, there is more inflation and unemployment in period one (in the long run), if it pools with rather than separates itself from the “wet”.
3. The “wet” approach in a separating and a pooling equilibrium: The second observation concerning the “dry” is not true for the “wet”. The “wet” clearly benefits from pooling.
Finally, the question arises as to what welfare implications exist in this example? In analyzing these implications we have to specify the intertemporal utility function of the trade union, V, since, until now, only its utility function for the first period, V1, matters. In analogy to the specification of the intertemporal utility function of the central bank, we set V=V0+V1 and define V0 symmetrically to V1. Further, we will compare the two equilibria calculated above with the equilibrium that would emerge with complete information. This latter equilibrium is described by the equations (7a) to (7c) and is characterized by e0=ρ0=q0=e1=0 and w1=-ρ1=q1/(1-a)=b in the case of the “dry” central bank (ρf=0, c2=0), and by e0=ρ0=q0=0 and w1=e1/d=q1/(a*d+1-d)=ρ1/(d-1)=b/(1-d) in case of the “wet” central bank (ρf=0, c2=d).
In line with the results above the following statements hold: (1) The trade union benefits the most in the separating equilibrium with the “dry”. It is indifferent, whether it is a separating equilibrium with the “wet”, or any equilibrium with complete information. It likes least the pooling equilibrium with the “dry” or the “wet”. The reason is intuitively explainable. What it wants most--because of the specification of its utility function--is a real appreciation of the currency by b in both periods. This is exactly what happens in the separating equilibrium with the “dry”. Next, it wants a real appreciation by b at least for one period which the trade union gets in an equilibrium with complete information or in a separating equilibrium with the “wet”. Finally, the pooling equilibrium is ranked last, since there the real appreciation lasts only for one period and is either too large (endangering employment too much) or too small (increasing real wages too little). (2) The welfare of the central bank depends on the size of a, the relative openness of the economy. It is interesting to note that if a<1/3 then the “dry” strictly does its best in the separating equilibrium with incomplete information, even better than in the case of complete information. If a>1/3 then the “dry” does better in the separating than in the pooling equilibrium the larger z, the probability the trade union attaches for having a “wet” as central bank.
Note that a “dry” and the trade union can improve their welfare positions strictly when moving from an equilibrium with complete to one with incomplete information (if it is a separating one with a<1/3). This outcome of a strict Pareto-improvement when information becomes incomplete is contrary to intuition, but results from inefficiencies of the equilibrium solution with complete information, 1/
1. Refinement of equilibria
Since off-equilibrium beliefs play a crucial role for the existence of the pooling equilibrium, in the perspective of the paper by Cho-Kreps, 1987, one could ask which refinements of equilibria are possible.
The Important point to note is that if equation (10b) holds then, for the “wet”, the message e0=0 dominates the message e0=x<0 since the “wet’s” utility is higher when choosing e0=0 whatever the response of the trade union is. Hence, an “elimination of type-message pairs by dominance” (Cho-Kreps, 1987, pp 199) is possible. Given the elimination of the strategy e0=x for the “wet”, the union knows for sure that when a central bank is sending the message e0=x, it must be the “dry”. Consequently, the off-equilibrium beliefs w1|y=z and w1|y=1, if e0=x<0 is sent, are no longer justifiable. The union should have a wage reaction w1|y=0, if the message e0=x<0 is sent. But then, the pooling equilibrium exists only if for the “dry” the following inequality holds
Otherwise there is a separating equilibrium.
2. Initial real misalignment
As mentioned above, in this section we deal with the question of an initial real misalignment. In such circumstances, the central bank will want to start an exchange rate commitment by setting e0= c1*ρf, with c1 appropriately defined according to the type of central bank.
The only difficulty which now arises is that the “dry” and the “wet” central bank react differently to an initial real misalignment due to the fact that the value of c1 is smaller for the “dry” than for the “wet”. Nevertheless, it is still attractive for the “dry” to keep exchange rates constant whatever the level of c1 (due to c2=0). For the “wet” to mimic the “dry” we can define the strategy space for the signaling game by e(o)=c1*ρf (c1 being the value of the “dry”) and x<c1*ρf so long as the difference between the c1-parameters does not become too important. Then all results still hold except for the interpretation of x which is being now positive (if 0<x<c1ρf) or negative, and therefore includes the possibility of a depreciation of the currency at the start of an exchange rate commitment, even if one plays tough as in the case of a “dry” in the separating equilibrium.
Suppose a central bank has changed its preferences from a “wet” to a “dry” now aiming at an exchange rate commitment to an inflation-free currency, e.g., to the DM. How does it want to signal this change to the public? The answer, according to this paper, is that it may pay off for the central bank to signal its newly adopted “dryness” by an initial exchange rate appreciation, even vis-a-vis the DM. This would be especially true, if z is large, the probability of still being regarded as a “wet” central bank by the public, i.e. the trade union.
Of course, the result depends on the reaction of domestic prices or wages to changes in the international competitiveness of a country (as assumed in equation (4)). If wages do not react to such changes, then any efforts to signal a newly adopted “dryness” by an initial real appreciation are not only In vain, but may significantly endanger the current account position of a country or its level of employment. Yet, if wages do react to such changes, and this seems to be more and more the case in Europe, given the competitive strength of the German economy, then there are benefits in signaling “dryness” as described above.
BackusD. and J.Driffill (1985) “Inflation and Reputation”American Economic Review 75 pp. 530-38.
BarroR. and D.Gordon (1983) “Rules, Discretion and Reputation in a Model of Monetary Policy,”Journal of Monetary Economics 12 pp. 101-21.
ChoI.-K. and D.M.Kreps (1987) “Signaling Games and Stable Equilibria,”Quarterly Journal of Economics 102 pp. 179-221.
GiavazziF. and M.Pagano (1988) “The Advantage of Tying One’s Hands. EMS Discipline and Central Bank Credibility,”European Economic Review 32 pp. 1055-75.
HochreiterE. and G.Winckler (1991) “The Appreciation of the Austrian Schilling Between 1979 and 1981: The Importance of Signaling” (forthcoming).
ObstfeldM.(1988) “Comments” to Giavazzi-Pagano (1988) European Economic Review 32 pp. 1077-82.
RogoffK.(1985) “The Optimal Degree of Commitment to an Intermediate Target,”Quarterly Journal of Economics 100 pp. 1169-90.
The author is a professor at the University of Vienna. This paper was written while he was a visiting scholar in the Research Department of the International Monetary Fund.
See also Obstfeld (1988), pp 1078.
Which signal conveys what message to the public is, of course, an open question. In some cases it may be sufficient to enter a fixed exchange rate relationship for setting such a serious signal.
If the starting point is already way out of line, i.e., if at the start there is massive unemployment or a large current account deficit because of an initial lack of international competitiveness, a depreciation of the currency may be unavoidable. Nevertheless the premise still holds that it pays off to signal toughness at the beginning of an exchange rate commitment.
What concerns current account deficits when an initial real appreciation occurs note that in our model the central bank only decides about exchange rates, so that the money supply or interest rates can be determined endogenously and, hence, the financing of the current account deficits via capital imports becomes a feasible option.
The analysis does not include empirical issues. The Austrian experience, when appreciating the Schilling between 1979 and 1981, will be discussed in a forthcoming paper by Hochreiter-Winckler (1991).
There is another way of interpreting the model: in Giavazzi-Pagano (1988), pp. 1058. e+p* is the price of the tradable, and p the price of the non-tradable good. The producers in the tradable goods sector are price takers, whereas the price of the non-tradable good is set with a fixed markup over wages. If wages increase, the price of the latter will increase at the same rate, while the price of the former remains constant thus making its production less attractive and thereby causing the current account and the employment level to deteriorate.
For a discussion of such a type of utility function see Backus-Driffill (1985). In our analysis we assume that ρ ≤ ρf.
A similar result is found by Vickers (1986), pp. 451.