THE EFFICACY OF CENTRAL BANK INTERVENTION ON THE FOREIGN EXCHANGE MARKET: UGANDA'S EXPERIENCE

Abstract

This paper employs conditional probabilities generated from a homogenous two-state Markov chain to obtain maximum likelihood estimates of the efficacy of central bank intervention on the foreign exchange market in Uganda. This enables us to explicitly model the fact that intervention actions usually do not target the exchange rate itself but rather its orderly movement. The results suggest that seasonal pressures are largely responsible for moving the short-term exchange rate process between the different state spaces. Intervention reduces the probability of the exchange rate process staying in a regime characterised by sharp and disruptive tendencies. Copyright © 2010 John Wiley & Sons, Ltd.

Full text

Page 1

The Efficacy of Central Bank Intervention on the Foreign Exchange Market:
Uganda’s Experience

















By

Eria Hisali

Faculty of Economics and Management
Makerere University, Uganda

mhisali@yahoo.co.uk/ehisali@fema.mak.ac.ug














A Paper submitted to the Centre for the Study of African Economies, Oxford
University

March, 2008

Page 2

Abstract

This paper seeks to contribute to understanding of the efficacy of central bank
intervention on the foreign exchange market by explicitly modelling the fact that
intervention actions usually do not target the exchange rate itself but rather its orderly
movement. Conditional probabilities generated from a homogenous two-state Markov
chain are employed to obtain maximum likelihood estimates of possible factors that may
govern a shift from one exchange rate state space to another. This enables us to find
answers to some pertinent questions such as: what are the possible factors that may
govern the shift from one state space to another in the short term? Is it possible that
central bank intervention is one of these factors, that is to say, is central bank intervention
effective? In other words, if for some reason it happens that large and disruptive
movements arise in the spot exchange rate process, what is the likelihood that central
bank intervention will succeed in returning it to a more tranquil regime? What would
happen on the other hand, if for any reason the bank intervened when the exchange rate
was in a tranquil regime?

The empirical results suggest that seasonal pressures are largely responsible for moving
the short term exchange rate process between the different state spaces. Intervention
reduces the probability of the exchange rate process staying in a regime characterized by
sharp and disruptive tendencies. This is an interesting result especially if seen in light of
the fact that interventions are small in relation to the overall size of the daily market
turnover. The results appear to suggest that what matters is the message sent out by the
interventions, which favours the signaling channel of intervention.

2

Page 3

1

Intervention typically entails buying and selling of a foreign currency against the
Introduction
domestic currency based on market conditions. Intervention operations usually seek to
ensure stability of the foreign exchange rate market (Rankin, 1998). Highly variable
exchange rates increase exchange rate risk and are likely to impede long term planning
and investment (Arize, Osang and Slottje, 2004). Evidence on the effectiveness of
intervention is generally mixed (Sarno and Taylor, 2001) but clearly, unsuccessful
intervention is costly. The existing literature seems to pay proportionately less attention
to possibility of an endogeneity bias and nonlinearity in the effects of intervention and
this perhaps explains the failure to resolve the debate on the efficacy of intervention.

The approach developed in Baillie and Osterberg (1997a, 1997b, 2000b) and Dominguez
(1998) works under assumptions of policy exogeneity of intervention and linearity in the
effects of interventions and abstracts from the fact that interventions represent a highly
irregular activity with prolonged intervals of inaction. These are strong assumptions. The
probability that the bank will decide to participate in the market and the probability that
the exchange rate will respond in the desired manner are both endogenous functions of
existing market conditions. Also, actual interventions are in practice undertaken
irregularly with some intervals of inaction, possibly because intervention within a small
neighborhood of equilibrium may result in a greater probability of instability (Taylor,
2004). In addition, in floating exchange rate arrangements banks do not seek to target the
level of the exchange rate itself but rather to mitigate the effects of short term disruptions
and return the exchange rate process to an orderly path. In other words central banks seek

3

Page 4

to move the exchange rate process to an orderly path without necessarily targeting the
level of the exchange rate itself.

This paper seeks to analyse the efficacy of central bank intervention on the foreign
exchange market. It employs conditional probabilities generated from a homogenous
two-state Markov chain to obtain maximum likelihood estimates of possible factors that
may govern a shift from one exchange rate state space to another. This enables us to find
answers to some pertinent questions such as: what are the possible factors that may
govern the shift from one state space to another? Is it possible that central bank
intervention is one of these factors, that is to say, is central bank intervention effective?
In other words, if for some reason it happens that large and disruptive movements arise in
the spot exchange rate process, what is the likelihood that central bank intervention will
succeed in returning it to a more tranquil regime? What would happen on the other hand,
if for any reason the bank intervened when the exchange rate was in a tranquil regime?

The approaches employed by Taylor (2004) and Beine, Laurent and Lecourt (2001) use
regime switching specifications to take account of nonlinearity and policy endogeneity of
intervention and to generate transition probabilities. An implicit assumption of their
frameworks is that determinants of the exchange rate are clearly known to the analyst.
This is a questionable assumption especially when seen in light of the controversy
surrounding the predictive ability of structural models of exchange rate determination. On
the other hand, specifications based on duration dependence (such as Taylor, 2004)
assume that markets clearly understand the different regimes and models governing the

4

Page 5

exchange rate process. The transition probabilities employed in this paper are initially
assumed to be generated by a homogenous Markov chain process. Subsequently then,
possible factors that may influence switching of the process from one state space to
another are explored. This approach therefore avoids the controversies surrounding
specification of the exchange rate models as functions of fundamentals or even duration
dependence in the process of generating the transition probabilities. This approach seems
plausible since most central banks do not target the exchange rate itself but rather seek to
mitigate the effects of short term disruptions and return the exchange rate process to an
orderly path.

An important innovation of this paper therefore is to explicitly take into consideration the
fact that intervention actions usually do not target the exchange rate itself but rather its
orderly movement in assessing the efficacy of intervention. This should provide a better
way to understand the effects of intervention. At the same time though, we make
recognition of the possibility that since banks can not a priori be sure of what the effect of
its actions will be, it can only intervene on a probabilistic basis; that its actions will have
the desired impact1.

The empirical results suggest that seasonal pressures are largely responsible for moving
the short term exchange rate process between the different state spaces. Central bank
intervention actions reduce the probability of staying in the regime characterized by sharp
and disruptive tendencies. This is an interesting result especially if seen in light of the

1 This emanates from the fact the influences that cause the disruptions are exogenous and outside of what
the bank can control and also the fact that there is a lot of controversy surrounding the efficacy of
intervention.

5