Post-Louvre Intervention : Did Target Zones Stabilize the Dollar?

Abstract

At their Louvre meeting in February 1987, the Group of Seven (G7) countries agreed to stabilize dollar exchange rates.Over the next two years, central banks frequently bought and sold dollars in a manner broadly consistent with attempting to maintain target zones, and dollar exchange rates appeared more stable than they previously had been.This paper investigates claims that the G3 (Germany, Japan, and the United States) successfully maintained target zones following the Louvre meeting.We use daily, official intervention data and simultaneous-equation techniques to estimate Probit reaction functions and GARCH exchange-rate equations.From the reaction functions, which include variables for target zones and market disorder, we construct Mill's ratios to serve as instruments for intervention.We introduce the Mill's ratios into both the conditional mean and conditional variance of the exchange-rate equations.The results suggest that the G3 reacted to exchange-rate movements in a manner broadly consistent with maintaining target zones.With some notable exceptions, however, we do not find strong evidence that the intervention successfully influenced subsequent exchange-rate movements.clevelandfed.org/research/workpaper/index.cfmattempt to maintain a target zone for both the mark-dollar and yen-dollar exchange rates, but that official dollar interventions generally failed to alter the direction of errant exchange rates.This could explain why the Group of Three (G3) seemed to abandon target zones by late 1988.Nevertheless, some of the results suggest that although intervention did not reverse exchange-rate movements, it may have smoothed them.We also find that intervention had mixed effects on exchange-rate volatility over the post-Louvre period.Section I of this paper presents models for the Probit reaction functions and the GARCH exchange-rate equations.We estimate these models in section 11.Here we discuss the construction of variables, notably our target estimates, and we describe the results in a fairly technical sense.In section 111, we relate our findings to other research in this area.I. h e Model Reaction function Central banks buy and sell foreign exchange to influence trend movements in exchange rates, to calm disorderly markets, to alter their reserve holdings, and to fulfill a variety of customer transactions.We define intervention as official transactions to influence spot exchange rates and therefore consider only the first two of these motives.Intervention to affect the trend movements in exchange rates could include "leaning against (or with) the wind" or maintaining target zones.Intervention to calm disorderly markets lacks a precise definition, but officials generally identify disorderly markets in terms of short-term volatility.Our definition of intervention conforms to what Adams and Henderson (1983) classify as clevelandfed.org/research/workpaper/index.cfm"activen intervention.We model post-louvre intervention as a function of a triggering mechanism that initiates intervention to alter trend movements in the exchange rate and as a function of a vector of variables that attempts to capture disorderly market conditions: B B (la) zF,t = 70 + T i ,,TI + D?, tri B + ei,t, (lb) ~s , t = r o + ~s , t r s + ~s , t r ; + ~s , t .In our model, the superscripts B and S refer to intervention purchases and sales of U.S. dollars, respectively, and the subscript i (-1,2) indexes the intervening country.In equation (la) , z;, is intervention purchases of dollars by the ith central bank at time period t; T : is the corresponding target-zone triggering mechanism at time t, and D: is a time-t vector of terms that defines disorderly markets.Corresponding definitions apply to equation (lb).All central banks react to the same trigger mechanisms and to the same measures of disorderly markets.We also assume thatand As we will explain in section 11, we include the lamed conditional variance from the exchange-rate equations as a measure of disorderly markets [see equations ( 7) and ( 14)l.clevelandfed.org/research/workpaper/index.cfmExchan~e-rate process A substantial body of literature suggests that exchange rates follow a martingale process with a heteroscedastic error term.The GARCH model initially proposed by Bollerslev (1986) is particularly well suited to variables exhibiting such behavior.The GARCH framework is also conducive to the study of intervention because it allows us to estimate simultaneously a conditional-mean equation and a conditional-variance equation, both of which can accommodate intervention terms on the right-hand side.We interpret intervention in the conditional-mean equation as measuring the impact of intervention on trend exchange-rate movements, and we regard intervention in the conditional-variance equation as measuring the ability of intervention to calm disorderly markets.Our exchange-rate model is (5) AS, = Xt6i + Z,62 + e,, (6) etlnt-lt(O,h,,v), In equation (5), AS, is the log change in the spot exchange rate at time t from time t-1; Xt is a vector of exogenous variables, and zt is a vector , of domestic and foreign intervention variables, such that B I (8) zt = rztt1 zzpt1 zttI ZLI.Equation (6) indicates that the distribution of the error term is conditional on information available at time t-1.By modeling the errors with a student-t distribution, we assume that the distribution is symmetric, but allow that it Bollerslev extended previous work by Engle (1982).clevelandfed.org/research/workpaper/index.cfmclevelandfed.org/research/workpaper/index.cfmTABLE 1 Apparent Exchange-Rate Target Ranges (foreign currency units per dollar) I.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 34 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.