PIERRE-GUILLAUMEMI=ON Universit~ Robert Schuman Strasbourg, France
A Model of Exchange Rate Crises with Partisan Governments* This paper investigates the consequencesof electionson the capacityof governmentsto defend a fixedparity in the presence of output shocks. It demonstrates that the politicaluncertainty associated with elections may significantlymodifythe reactionsof governmentsin a way that may be interpreted as opposite to their ideology. It also shows that the probabilitythat a devaluation occurs increases significantlyafter an election. 1. I n t r o d u c t i o n The influence of political factors on economic policy has been considered crucial for a long time. Considerable time and effort have been devoted to the observation and understanding of the link between economic and. institutional variables. Exchange rate regimes are no exception to the rule and two puzzling stylized facts have been repeatedly observed. First, as Milesi-Ferretti (1995) underlines, governments often choose an exchange rate regime that does not seem in accordance with their political stances. Thus, it is usually assumed that left-wing igovernments are more averse to output fluctuations and less sensitive to inflation than right-wing governments. Left-wing governments should therefore favor flexible exchange rate regimes to stabilize output, thanks to an active monetary policy. That view does not coincide with the fact that it was socialist administrations that joined the EMS in the eighties, both in France and Spain, and that, on the other hand, it was a conservative government that ended the participation of the U.K. in the EMS in the nineties. More rigorously, Eichengreen, Rose and Wyplosz (1995) provide a statistical analysis of episodes of turbulence in foreign exchange markets on a sample of twenty O E C D countries between 1959 and 1993. They found no significant correlation between the political orientations of governments and the occurrence of such episodes, be it devaluations, revaluations, flotations or failed specttlative attacks. The second striking stylized fact about the operation of fixed exchange rate commitments is that they are very sensitive to elections. More precisely,
*Helpfulcommentsby two anonymousrefereesand by seminarparticipants at the University Robert Schuman of Strasbourg and at the UniversityLouisPasteur of Strasbourgare gratefully acknowledged.E-mail:pierre-guillaume.meon@iep_u-strasbg.fr. ]aurnal of Macroeconomics, Fall 2001, Vol. 23, No. 4, pp. 517-535
Copyright© 2001 by LouisianaState UniversityPress 0164-0704/2001/$1.50
517
Pierre-Guillaume M~on
devaluations tend to be preceded by political instability. For instance, Edwards (1994) found out that in Latin American presidential democracies, 77% of the devaluations took place within the first eighteen months of the government being in office. 1 His finding is corroborated by Klein and Marion (1997) who observed that the probability of a devaluation increased significantly in the month that follows an executive transfer in sixteen Latin American countries between 1956 and 1991. Eichengreen, Rose and Wyplosz (1995) obtained similar results for twenty OECD countries, thereby proving that the relationship is not peculiar to developing countries. Nonetheless, whereas the empirical relationship is well documented, surprisingly little work has been devoted to date to giving those facts a sound theoretical explanation. Most authors get away with assuming that the political cost of a devaluation is smaller in the morrow of an election. Although there must be some truth in that argument, it does not explain why governments should want to devalue in the first place nor does it explain the partisan patterns underlined by the first stylized fact. From that point of view, Milesi-Ferretti (1995) provides an illuminating but unfortunately isolated argument. Building on Barro and Gordon (1983b), Milesi-Ferretti (1995) shows that an inflation-averse right-wing incumbent may find it profitable to adopt a floating exchange rate so as to promote its stronger monetary credential in the eyes of voters. On the contrary, an inflation-prone left-wing incumbent may adopt fixed exchange rates in order to hide its weakness. That argument gives a most convincing account of the first stylized fact. It is however hard to reconcile it with the second stylized fact. The present paper investigates further the impact of elections on the viability of a fixed exchange rate regime when monetary policy is steered by partisan governments. It builds on the escape clause model of exchange rate crises that originates in Flood and Isard (1989) and Lohmann (1990) and was adapted to study exchange rate crises by Obstfeld (1994) or Andersen (1998). It is therefore an extension of two strands of the literature, devoted respectively to exchange rate crises and to the flexibility versus credibility trade-off. The present analysis shows that the political uncertainty associated with elections may be the source of output shocks originating in the wage setting behavior of the private sector. It also shows that in polarized political systems, elections may even preclude the capacity of inflation-averse governments to defend a fixed parity. The paper is organized as follows. The first section describes the model that allows us to study the credibility of a fixed parity. The second section 1That proportion still amounted to 70% in parliamentary democracies and 43% in undemocratic regimes.
518
A Model of Exchange Rate Crises
shows that such a commitment is contingent on the realization of output shocks. The third section shows that once expectations are taken into account, elections alter the exchange rate policy of governments and that they can even hasten the collapse of fixed exchange rate commitments.
2. The Modol To describe the supply side of our model, we assume that output is a decreasing function of the real wage. Letting Yt, w t a n d Pt denote the logarithms of national income, of the real wage and of the price level, respectively, the supply function consequently reads y, =
--(w~
-
p,) + u , ,
(1)
where u t is a stochastic shock, the mean and variance of which are, respectively, zero and o-~. Wage contracts are signed at the beginning of each period and last until the end of it. As private agents are rational and wish to preserve their purchasing power, they claim a nominal wage that covers the variations of the price level such that we = E~-l(pt) -
(2)
We moreover suppose that purchasing power parity holds. Ifet denotes the logarithm of the price of one unit of the foreign currency expressed in t e r m s o f the national currency and p* measures the logarithm of the foreign price level we can write et = Pt -
pt* .
(3)
In a small open economy, we can consider that the foreign price level is independent of national policies and domestic shocks. It can therefore be assumed constant. 2 Moreover, we can normalize units of foreign goods so that the logarithm of the foreign price level equals zero. This specification allows us to identify the price of the foreign currency with the domestic price level, and Equation (1) can be rewritten as an augmented Phillips curve, where only unanticipated exchange rate swings produce real effects: 2As the foreign monetary authority is only interested in stabili~ng foreign shocks,u t may be interpreted as the difference between domestic and foreign shocks, which is not absorbed by exchange rate pegging. 519
Pierre-Guillaume Mdon yt = ot(et -
e~) + u t ,
(4)
where e7 -= Et-l(et). To model the demand side, we suppose that the price level is an increasing function of the money stock. If we moreover assume that the authorities control the money stock, we can think directly in terms of the exchange rate. The authorities are averse both to inflation and to the fluctuations of output around an optimal level g which is strictly greater than the natural level of output. We moreover assume that domestic policy-makers face a fixed cost c of realignment. That cost may encompass the various inefficiencies arising in a flexible exchange rate regime or after a devaluation. It may also reflect the political and diplomatic costs of reneging on a promise to fix the exchange rate. It may finally take into account the loss of credibility or the sanction imposed on the policy-maker by the private sector, as in Barro and Gordon (1983a). To take those elements into account, we describe the authorities' utility by a standard quadratic loss function to which a constant term is appended when the parity is realigned: 1
l, = ~ (p~ -
01
p~_l) 2 + ~ (y~ -
9) 2 + ~ ,
(5)
where ~t is a dummy variable which is equal to One whenever et # et-t and equals zero otherwise. 0~ is a parameter that represents the aversion of the authorities for output relative to inflation fluctuations. That parameter reflects the authorities' "ideology" and is determined by an election in which two parties run. a Those parties, named party L and party R, exhibit similar loss functions but differ in the weight they attach to the cost of output fluctuations. Typically, party R is more conservative than party L, which means that 0L > 0a. As in Alesina (1987), both parties have a positive likelihood to win the ballot which amounts to p for party L and (l - p) for party R. To complete our description of the model, we assume the following tSming of events: First, agents sign their wage contracts, which are binding until the end of the period. Second, elections take place and the identity of the government is revealed. Third, the actual value of the disturbance term ut is realized. Fourth, the authorities set the exchange rate. Figure 1 below 3We use the term election for convenience here. It must nevertheless be stressed that our set-up is consistent with any kind of executive transfer, be it regular or irregular (i.e., coups).
520
A Model of Exchange Rate Cr~ses Contracts wt
Election
Shock
Monetary policy
0,
ut
et
l
[ Period t
I
• time
Figure 1. The Timing of Monetary Policy
summarizes that scenario. 4 The question is now to understand under which circumstances the authorities respond to the shock by readjusting the exchange rate and when they defend a fixed parity.
3. A Fixed Exchange Rate as a Contingent C o m m i t m e n t
The authorities decide to adjust the exchange rate whenever such a behavior allows them to reduce their losses. To see when that condition is met, we must therefore first determine the authorities' losses when they adjust the exchange and when they do not. When they realign, the authorities set the exchange rate so as to minimize their loss function (5). Substituting the Phillips curve (4) into the objective function (5) and minimizing for a given level of the expectations yields the authorities' reaction function: eTdj _
1
1 + 0ia 2 let-1 + 0,a2e7 - 01c~(ut - g ) ] .
(6)
Plugging the above exchange rate in the Phillips curve (4), one obtains the discretionary level of out-put, taking expectations as given: y~J -
1 [~(et-i 1 + 0ia 2
-
e~) + 0,a~# + ut] .
(7)
4This timing is a convenient way to model that nominal wage adjustment is more sluggish than exchange rate adjustment.
521
Pierre-Guillaume M g o n
From (6), (7) and (5) we find the authorities" losses when they realign: l ~ = 02.
1
[et(e~ - et-1) + g - ut] z q- c
2 1 + O~a2
(8)
If, facing the same shock and the same expectations, they decided to defend a fixed exchange rate e t = et_l, t h e Phillips curve indicates that they would obtain the following level of output: y~
=
-a(e7
-
et-1) + u t .
(9)
The corresponding losses would therefore amount to Oi
(10)
We can now specify the circumstances under which the authorities decide to realign. They do so whenever l ~ j < / ~ . From (8) and (10) that condition is equivalent to 0/2a2 (ae7 - aet_ 1 + g - ut) 2 > 2c 1 + 0~az
(11)
Expression (11) shows that the decision to adjust the exchange rate depends on the realization of the shock. The peg is abandoned whenever the absolute value of ut is large enough. Treating (11) as an equality and soIving for its two roots, one finds the trigger values of ut for which the authorities decide to adjust the exchange rate parity:
~,
=
~(e~
-
~_~)
+
g -
1 .,/2(1 + O,a2)c
(12a)
1 ~/2(1 + OioLZ)c.
(mb)
0--fi
and 61 = a(e~ -- e t - i ) + g +
Accordingly, the authorities find it optimal to defend the announced exchange rate whenever ut ~ [ut, ai]- On the other hand, they devalue whenever ut < u_, and revalue whenever ut > 6iFrom (12a) and (12b), it is easy to notice that the critical values ofut are symmetric around a constant. However, that constant depends on the 522
A Model of Exchange Rate Crises
private sector's expecta~ons. The role of those expectations is crucial when elections are taken into account. We must therefore describe them more precisely.
4. The Sustainability of a Fixed Parity in the Presence of Elections In this section, we describe explicitly the way market expectations are formed. We subsequently determine the formal expression of the shocks for which the authorities decide to revise the exchange rate parity once expectations are taken into account. We finally describe the behavior of governments as a f~mction of their ideologies and of the political system with which they are faced. The Formation of Market Expectations When private agents form their expectations about the exchange rate, they are confronted with two uncertainties. In the first place, they do not know the identity of the winner of the election. They must consequently expect an average of the exchange rates implemented by the two parties,_ weighted by their probabilities of winning the ballot. Accordingly, the private sector expects the following exchange rate:
e7 = p'E(etl0 = 0c) + (1 - p)'E(etlO = OR)-
(13)
However, private agents also face a second uncertainty which pertains to the,realization of the shock ut. They consequently do not know whether the elected government will find it optimal to stick to the initial parity or to adjust the exchange rate. The expected value of the exchange rate conditional on party i's being in office therefore reads: E(etlO = Or) = prob(~ < ut < a~)'et-1 + prob(ut < ~)-E(etlut < ui) + prob(ut > a~)-E(etlut > ~i~) .
(14)
To solve this equation, we have to specify the distribution of u t. We therefore suppose that ut is uniformly distributed on the interval [ - ~t; ~t].5 Provided ut is linearly distributed, we can find a simple expression of party i's expected reaction function. Some fairly cumbersome but straightforward algebra, which is reported in the appendix, establishes that SThis specification rules out multiple equilibria which arise in similar frameworks such as Obstfeld's (1996) or Ozkan and Sutherland's (1998).
523
Pierre-GuiIlaume Mgon E(etlO = 0~) = e~-i + 1 +0,aO~a2 (ae~ - aet-1 + if)
(15)
We can subsequently get an expression of the unconditional expected exchange rate by combining expressions (15) and (13), namely,
el = et-1 + ag
p0L + (1 -- p)0 R + a~OLOn 1 + a2[(1 -- p)0z + p0R]
(16)
It is noteworthy that the expected exchange rate is an increasing funct_ion of 0L, 0R and p. Indeed, the more sensitive the parties are to output fluctuations, the more likely they are to depreciate the currency. The likehhood of a devaluation is even higher when the probability that the more sensitive party is elected increases, hence the role of p. The Definition of Trigger Shocks in the Presence of Elections Now that we have a tractable expression of the private sector's expectations, we can specify each party's behavior if it is elected. To do that, we simply plug expression (16) in the expressions of the critical values ofut:
(1 + 0La2)(1 + 0na z)
1 ,/2(1 + 0ia2)c
(17a)
1 ,/2(1 + O,a2)c
(17b)
u__, = 1 + ~-z[-(1-~ p ) ~ + p0 a] # and (1 + 0La2)(1 + 0oa 2) a, = 1 +
+
g +
From the above expressions, we see that the boundaries of the interval depend upon structural parameters, such as the target level of income and the slope of the Phillips curve, but also on subjective variables, such as the parties' aversions for output fluctuations. It is noteworthy that the government's reaction depends not only on its own preferences but also on its opponent's and on their respective ex ante probabilities of winning the ballot_ Above all, it should be noted that the boundaries of interval [u~, at] are independent of parameter ~t. Accordingly, the interval of tolerable shocks may or may not be included in the distribution of possible shocks [ - ~t; ~t]. Observation of expressions (17a) and (17b) reveals that if the target level of output ff is large, if parties are highly averse to output fluctuations, that is, 0L and OR are large, and output volalSlity is moderate, that is, [ - rt; ~t] is narrow, then at least ai if not ui may well be greater than ~t. Under 524
A Model of Exchange Rate Crises those circumstances, [ui, a~] is not included in [-~t; ~t]. On the other hand, ffthe target level of output is small, ffparlSes are moderately averse to output fluctuations and output volatility is high, the boundaries of [ui, al] fall within those of [ - ~t; ~t]. For simplicity sake, we will henceforth refer to the situation where ~i, a~] C [ - ~t; ~t] as the conservative political system case and to the situation where a~ > ~t as the nonconservative system case. We use the adjective conservative as defined in Rogoff (1985) because, ceteris paribus, at least one party is more sensitive to output fluctuations in the latter configuration than in the former. 6 As it will be made clear below, those two configurations lead to qualitatively different results. In what follows, we will therefore treat them separately. In particular, we will try to assess the impact of elections on the probability that the exchange rate is defended and on the magnitude of the shocks that trigger a reaction of the government.
The Reactions of Governments in a Conservative Political System Further examination of expressions (17a) and (17b) allows us to be more specific about the incidence of elections on the behavior of both parties when they are elected in a conservative pohtical system. We can first compute the length of the interval within which the government can tolerate output shocks without adjusting the exchange rate, which is simply the length of the interval [ui, ad:
2 ,/2(1 +
ui - u--i - Oicx
OioLZ)c.
(18)
On the other hand, if we focus on the midpoint of the interval, and denote it u~ec, we can write
uet~c
(1 + 0La2)(1 + e r a 2) 1 + a2[(1 -- p)OL + P0n] ~]"
(19)
From (18), it is plain to see that the range of tolerable shocks is an increasing function of the cost of breaking the commitment. Most of all, it is decreasing in the government's concern for output fluctuations. 7 6Theoretically,a third configurationcould arise where [ -I~; g] C [_Yi,_ fh], for instance, if the cost of leaving the peg is prohibi~veor if the volatilityof output shocks is very small. In this configuration the governmentsystematicallydefends the peg and the questionwe address in this paper becomes a non-issue.We will therefore rule it out. 7Thus, a(al - ui) _ -2c(2 + 0# 2) < 0.
ao,
o~,~,/~(1 + o,,~~) 525
Pierre-Guillaume Mdon From (19), it appears u*t,c is strictly positive, which is a familiar result of the literature, s Strikingly, it is also identical for both parties. Moreover it is a function of their preferences and of their ex ante likelihood of winning the election. This result reflects the impact of electoral uncertainty on private agents' expectations. Like the expected exchange rate, the midpoint of interval [ui, ai] is an increasing function of both 0L and 0a 9 and is also increasing in p.10 This means that a government's reaction to output shocks does not only depend on its own preferences but also on the preferences of the opposition and on their respective ex ante likelihoods of winning the poll. In the case of a conservative system, tolerable shocks belong to the interval of possible shocks [ - g ; g]. There is therefore a direct relationship between the length of the interval [u~, ai] and the probability that the exchange rate is defended. Our first proposition is then straightforward. PROPOSITION 1. In a conservative political system, the range of tolerable shocks and the probability that the exchange rate is defended are decreasing functions of the government's aversion for outpnt fluctuations. It is moreover independent of the opposition's preferences.
Proof.
From expression (18) and footnote 7, we know that 0(lil - ui)/00i <0. Besides, expression (18) only features the government's relative aversion to output fluctuations, 0i. i In other words, proposition 1 means that the less conservative the government, the more sensitive it is to output shocks and the less likely it is to defend the peg. Consequently, party L should be more sensitive to shocks than party R. However, in a conservative system, political uncertainty per se does not impinge on the probability that the peg is defended. Nevertheless this result does not mean that elections have no effect in a conservative political system and this is the point of our second proposition. 8The llnding that the interval of tolerable shocks is not symmetricaround zero also appears in Lohmann(1990),Andersen (1998)or MeNus (1999). paz (1 + oRaz) 9Thus' 0u*u~ _ [1 + ct e [(1 - p)0L + pORJ]z !7 > 0 and 00L (1 - p) a 2 ( 1 + 0haz) [1 + a2[(1 - p)0L + p0a]]z9 > 0. 0% lOlndeed' 0u% a s (0n - %)(1 + %a~)(1 + 0actz) - -0p = F ~= ~~-~ z p~-~T pO~]] ~ 9
526
> o. as % > %
A Model of Exchange Rate Crises PROPOSITION 2. In a conservative political system, elections alter the reaction of the government to defend an exchange rate peg in the presence of output shocks. The party that is the more (least) averse to output fluctuations tends to tolerate greater (smaller) negative shocks and smaller (greater) positive shocks than it would in the absence of electoral uncertainty.
Proof.
Let us denote u*, party i's interval midpoint in the absence of electoral uncertainty, that is, when the probability that it wins the poll is strictly equal to one. If we recall from footnote 10 that u*~ is an increasing function of p, we can write
u~ < u~,c < u ~ , with
(20)
u~* = (1 + 0ia2)9.
To understand the true meaning of that result, we should compare each party's reaction with what it would be in the absence of elections. If we focus on party R, expression (20) reveals that in normal times that party tolerates important negative (contractionary) shocks but only moderate positive (expansionary) shocks. After an election, the set of tolerable shocks " moves upward, which explains why u~ec is greater than u~. The presence of elections alters party R's behavior insofar as it makes it more sensitive to negative shocks and less sensitive to positive ones. The rationale for that result pertains to the wage setting behavior of the private sector. During an electoral period, private agents expect an exchange rate that amounts to the weighted average of the exchange rates conditional on the success of party L or party R. As party L is more inflationprone than party R, the latter inherits a higher nominal wage than it would in the absence of electoral uncertainty. A higher nominal wage constitutes a recessionary shock that adds up to negative output shocks but compensates positive ones. A government of type R will therefore tolerate greater positive shocks after an election than in normal times. It is straightforward to see that the impact of electoral uncertainty on the behavior of party L is the opposite. Party L tolerates greater negative shocks and smaller positive shocks than it would without electoral uncertainty. In that case, elections allow that party to inherit a lower nominal wage than in normal times, In a sense, Party L benefits from its opponent's aversion for inflation, which explains why it is less tempted to react to adverse output shocks. Moreover, it should be stressed that the effect of elections is stronger, the less the elected party was likely to win the poll ex ante. A government's behavior will be closer to its behavior in normal times the more likely it was 527
Pierre-Guillaume M~on to come to power, n This result stems from the fact that the more likely a party is to come to power, the closer are the expectations it inherits to those it would face without political uncertainty. This result may provide an alternative complementary explanation to the first stylized fact. Thus, owing to the way expectations are formed, parties' behaviors may look counterintuitive ex post and governments may react to shocks which they would have disregarded in normal times. Another explanation may also be found in Persson and Svensson (1989) or Milesi-Ferretti (1996). In their frameworks, the convergence of government policies stems from a principal agent problem where the incumbent tries to influence its successor's policy through the ma_uipulation of a state variable. Our framework differs from theirs in several respects and in particular because it does not rely on a principal-agent model and because it describes explicitly the operation of an escape clause. Our findings so far rest on the restriction that the interval that contains tolerable shocks is included in the interval [ - ~t, ~t]. If we drop that restriction we may reach qualitatively different conclusions. We do so in the next subsection.
The Reactions of Governments in a Non-Conservative Political System By definition, in a non-conservative political system, we have ~, > g, which significantly modifies the analysis. We therefore reach different conclusions that are summarized in our third proposition. PROPOSITION 3. In a non-conservative political system, the probability that the exchange rate is defended may be very low. The probability that party R defends the parity becomes a decreasing function of party L's aversion for output fluctuations and of party L's ex ante probability to win the poll. In extreme situations the probability that the exchange rate is defended becomes strictly nil.
Proof The first part of the proposition follows from the fact that, according to (17a), Ou_i/80~ = ~u*~c/O0~ - ~ O(a~ - ui)/80~ > 0. Therefore, the set of tolerable shocks in a non-conservative system, [ul; g], shrinks when the sensitivity of the government to out-put fluctuations increases. The second part of the proposition stems from the fact that Ou_R/80L = Ou*~c/OOL > 0 and 0u_R/0p = 0u*~/0p > 0. There again, this means that the range of tolerable shocks for a government of type R shrinks whenever 0L or p increases. 11Thus,lirn u*~ = u*Land lirn u*z~ = w-~
528
W-'0
A Model of Exchange Rate Crises The final part of the proposition follows from the fact that 0L and 0h are unbounded. As lim u,. = co > ~t, we see that the set of tolerable shocks may be empty. 0~-~ The intuition behind the first part of the third proposition is similar to the one of Proposition 1. If the aversions of both parties to output fluctuations are high, then they are less likely to sacrifice that objective to the defense of the exchange rate peg. The second part of the proposition is peculiar to a non-conservative system and deserves more attention. In a non-conservative system, the inflationary bias is such that the rise in the nominal wage constitutes a contractionary shock that may only be compensated by very high positive shocks. As party L becomes increasingly concerned about output fluctuations, its inflationary bias increases and expected inflation therefore rises, which raises the nominal wage for both parties. The contractionary shock therefore increases, which raises the cost of defending a fixed parity even for a government of type R. Therefore, the magnitude of the output shock needed to compensate that contractionary shock increases. It consequently clearly appears that in a non-conservative system, the probability that the more conservative party defends a fixed parity is a decreasing function of its oppo-nent's aversion for output fluctuations. The last part of Proposition 3 stresses that in some situations the parity may never be defended. The first reason may be that both parties are too sensitive to output fluctuations. The inflationary bias and expected inflation may then be so high that the rise in the nominal wage may constitute so high a recessionary shock that even the most favorable output shock may not compensate it. A devaluation consequently becomes unavoidable. However, for this phenomenon to be observed it suffices that only party L's aversion for output fluctuations becomes very high. In such a polarized political system, party L's aversion for output fluctuations is so high that it is responsible for the adverse shock in the nominal wage, even though party R would be able to defend a fixed parity in the absence of an election. Party R is therefore forced out of the peg because of its opponent's radicalization. The national currency is then devalued regardless of the identity of the government. Our third proposition shows that the influence of elections on the sustainability of an exchange rate may be more pervasive than what our first propositions suggested. Indeed, political uncertainty may not only alter the type of shocks to which the government responds but also raise the probability that the exchange rate is devalued. This finding may help us understand why post electoral periods are particularly favorable to devaluations. After an election, two situations may 529
Pierre-Guillaume M~on arise. First, a party whose commitment to fixed parities is weak is elected. Such a party is very likely to devalue the currency. Second, a party whose commitment is stronger is elected. However, as that party suffers from its opponent's low credibility, it is also likely to devalue. After an election both types of government are thus likely to devalue. This is why the initial parity is abandoned with a higher probability than in normal times, which is consistent with the second stylized fact.
5. Concluding Remarks Commitments to defend a fixed parity are political decisions that can be changed in adverse circumstances. The defense of such a commitment therefore depends on the preferences of the authorities, on exogenous shocks and, insofar as they affect economic outcomes, on market expectat-ions. As elections may alter the preferences of the authorities and market expectations, they may bear crucial consequences on the viability of a fixed exchange rate commitment. The present analysis tackles this issue. It shows that in the period following an election, the survival of a fixed parity becomes a function of the preferences of the elected government but also of the preferences of the opposition, in addition to output shocks. The intuition behind this finding lies in the way expectations are formed. Prior to elections private agents face an electoral uncertainty and can only anticipate an average inflation rate when they sign their wage contracts. As wage contracts are binding, both partie s inherit the same nominal wage when they come to power: Consequently, the inflation-prone party benefits from its challenger's credibility whereas the inflation-averse party has to cope with a nominal wage that incorporates its opponent's inflationary bias. In a conservative political system, that mechanism may force the ruling party to adopt a counterintu_itive behavior. Thus, the inflation-averse party may tolerate higher expansionary shocks but smaller recessionary shocks. On the other hand, the inflation-prone party may defend the parity for greater recessionary shocks but smaller expansionary shocks. In a non-conservative political system elections may moreover raise the likelihood of a devaluation. In extreme cases, a devalualion may even occur with certainty. As with any model, the present model relies on assumptions that may raise criticism. We for instance focus on fixed election dates. It must nevertheless be underlined that the same mechanism may be at work during any period, in countries with variable electoral terms, as Ellis and Thoma (1991) have shown. That mechanism may therefore explain why political uncertainty in a country is positively correlated with the adoption of flexible exchange rates practices, as Edwards (1996) observes. 530
A Model of Exchange Rate Crises
Second, although our model underlines that devaluations should be more likely during the months that follow an election and until wage-setters have revised their wage contracts, it is a one-period model and does not allow us to be more specific about the exact liming of devaluations. A fruitful development would therefore be to extend our model to a dynamic setting following for instance Ellis and Thoma (1991) or M6on (1997). Third, in our model governments act in a purely partisan fashion and no attention is paid to their reelection efforts. Alternative mechanisms highlightiaag such opportunistic considerations could be put forward to account for the same stylized facts. We believe however that the two approaches should be seen as complements rather than substitutes. Moreover, a few complementary stylized facts tend to suggest that partisan considerations play a crucial role in the evolution of exchange-rate regimes. On the one hand, there is evidence of partisan cycles in exchange rates. For instance, GLrt-ner (1986) or Bachman (1992) show that political news concerning the identity of the government significantly alter either the spot rate or the forward bias. More recently, Blomberg and Hess (1997), Lobo and Tufte (1998) or Garfinkel, Glazer and Lee (1999) find evidence of ideological effects on the level and on the volatility of the exchange rate. On the other hand, Ellis and Thoma (1996) find evidence of pariSsan cycles in variables relevant tothe operation of exchange-rate regimes such as the current account, the real exchange rate and the terms of trade. Fourth, the present analysis rests on an expectations-augmented Phillips curve. Similar arguments could however be drawn from alternative mechanisms highlighting for instance the role of public debt or distributive considerations, as Obsffeld (1996) points out. Finally, an interesting extension would be to drop the small economy hypothesis so as to study the synchronization of political cycles in interdependent economies. Though interesting, those extensions would probably foster rather than qualify this paper's main results, which help explain the two stylized facts that are pointed out in the introduction. Above all, it is a reminder that political factors are of paramount importance when it comes to evaluating the viability of a fixed parity. For that matter, they are closely akin to the conclusions Einzig (1968, p. 103) reached thirty years ago that "in given circumstances political influences are liable to overshadow economic influences among the causes of foreign exchange crises." Received:October1999 Finalversion:January2001 References Alesina, Alberto. "Macroeconomic Policy m a Two Party System as a Repeated Game_'" Quarterly Journal of Economics (August 1987): 651-77.
531
Pierre-Guillaume M3on Alexius, Annika. "Inflation with Consistent Escape Clauses." European Ecorwmic Review 43 (March 1999): 509-23. Andersen, Torben M. "Shocks and the Variability of a Fixed Exchange Rate Commitment." Open Economies Review 9 (April 1998): 139-56. Bachman, Daniel. "The Effect of Political Risk on the Forward Exchange Bias: The Case of Elections."Journal of International Money and Finance 11 (April 1992): 208-19. Barro, Robert, and David Gordon. "Rules, Discretion and Reputation in a Model of Monetary Policy." Journal of Monetary Economics 12 (July 1983a): 101-22. • "A Positive Theory of Monetary Policy in a Natural Rate Model." Journal of Political Economy 91 (August 1983b): 589-610. Blomberg, S. Brock, and Gregory D. Hess. "Politics and Exchange Rate Forecasts," Journal of International Economics 43 (August 1997): 189-205• Edwards, Sebastian. "Trade Liberalization Reforms in Latin America." In Latin America's Economic Future, edited by G. Bird and A. Helwege. London: Academic Press, 1994• • "Exchange Rates and the Political Economy of Macroeconomic Discipline." American Economic Review 86 (May 1996): 159-63. Eichengreen, Barry, Andrew K. Rose, and Charles Wyplosz. "Exchange Rate Mayhem: the Antecedents and Aftermath of Speculative Attacks." Economic Policy 21 (October 1995): 249--312. Einzig, Paul. Foreign Exchange-Crises: An Essay in Economic Pathology. London: Mac Millan, 1968. Ellis, Christopher J., and Mark A. Thoma. "Partisan Effects in Economies with Variable Electoral Terms." JournaI of Money, Credit and Banking 4 (November 1991): 728--41. --. "The Implications for an Open Economy of Partisan Political Business Cycles: Theory and Evidence.'" European Journal of Political Economy 11 (December 1996): 635-51. Flood, Robert P., and Peter Isard. "~MonetaryPolicy Strategies." IMF Staff Papers 36 (September 1989): 612--32. Garfinkel, Michelle R., Amihai Glazer and Jaewoo Lee. "Election Surprises and Exchange Rate Uncertainty." Economics and Politics 11 (November 1999): 255--74. Gartner, Manfred. "Some Political Economy of Flexible Exchange Rates." European Journal of Political Economy 2 (May 1986): 153-68. Klein, Michael W., and Nancy P. Marion. "Explaining the Duration of Exchange-Rate Pegs." Journal of Development Economics 54 (December 1997): 387-404. 532
A Model of Exchange Rate Crises Lobo, Bento J., and David Tufte. "Exchange Rate Volatility: Does Politics Matter?" Journal of Macroeconomics 20 (Spring 1998): 351-65. Lohmann, Susanne. "Monetary Policy Strategies--A Correction: Comment on Flood and Isard." IMF Staff Papers 37 (June 1990): 612-32. M~on, Pierre-Guillaume. "Credibility of Fixed Parities when the LifeExpectancy of Policy-Makers is Curtailed by Political Uncertainty." In Geld, Finanzwirtschaft, Banken und Versicherungen 1996, edited by C. Hipp et al. Karlsruhe: Verlag Versichertmgswirtschaft, 1997. Milesi-Ferretti, Gian-Maria. "The Disadvantage of Tying Their Hands: On the Political Economy of Policy Commitments." Economic Journal 105 (November 1996): 1381-1402. Obsffeld, Maurice. "The Logic of Currency Crises." NBER Working Paper 4640, February 1994. • "Models of Currency Crises with Self-Fulfilling Features.'" European Economic Review 40 (April 1996): 1037-47. Ozkan, F. Gulcin, and Alan Sutherland. "A Currency Crisis Model with an Optimizing Policymaker." Journal of International Economics 44 (April 1998): 339-64. Persson, Torsten, and Lars E. O. Svensson. "Why a Stubborn Conservative Would Run a Deficit: Pohcy with Time-Inconsistent Preferences." Quar-. terly Journal of Economics 104 (May 1989): 325-45. Rogoff, Kenneth. "The Optimal Degree of Commitment to an Intermediate Monetary Target." Quarterly Journal of Economics 100 (November 1985): 1169-90.
Appendix To get an expression of party i's reaction function, we must recall that it will defend the initial parity as long as the realization of the shock ut falls within the boundaries of interval [ui, zit]- Dropping the party suffix for notational convenience, we can write
E(et) = prob(u_ < ut < a)-et_l + prob(ut < u)'E(etlut < u) + prob(ut > ~2).E(etlut > z~) .
(A1)
As ut follows a linear distribution, the conditional probabilities on ut can be written as ~2
-
prob(u < ut < 5) = - 2g
u_u_
;
(Aea)
533
Pierre-Guillaume MEon
u+g prob(ut < u ) =
prob(ut > a) -
2g
~ -
2g
(A2b)
'
(A2c)
a
Moreover, using expression (6) we can also rewrite the expected values of the exchange rate conditSonal on the realization ofut:
E(et~t
= u_ + ~t i + Ocx2
(N2 _ u 2) + (ae7 -
c~et-1 + 17)(u + la)]
(A3a)
and
E(etlut > ~) =
1 ~t
-
o~ a
[~(a2
g2)
i + 0a z +(aet
-- etet_l + 17)(/.t -
a)/._1
(A3b)
Combining expressions (A2a, b, c) and (A3a, b) with expression (A1) yields the following expression for the expected exchange rate:
e~ = et_ 1 + 1 + OeL~
4~t
+ (ae~ -
aet_l
+ g)
1 +
.
(A4)-
Plugging the expressions of u_i and at as given by (12a) and (12b) in (A4), the expression of the expected exchange rate boils down to 0OL
e7 = et_ 1 + 1 + 0a 2 (ae~ - aet-1 + !7).
(A5)
Variables Definitions yt
wt Pt p* 534
= = = =
logarithm logarithm logarithm logarithm
of national income for period t. of the real wage for period t. of the domestic price level for period t. of the foreign price level for period t.
A Model of Exchange Rate Crises = logarithm of the price of one unit of the foreign currency expressed in terms of the national currency for period t. = expected exchange rate for period t. et a:>0 = slope of the Phillips curve. ut ~ [-rt; ~t], ~t > 0 = period t's out-put shock. l, = the gnvemment's loss function. = party L and party R's aversions for output relative OL > On > 0 to inflalSon fluctuations. g = target level of output. c>0 = fixed cost of a realignment. = a dummy variable which is equal to one whenever e~ ~ et-1 and equals zero other~se. p>O = probability that party L wins the election. [~i, a,] = the set of shocks for which the government defends the exchange-rate peg. et
535