Currency Substitution in Laos

• Author: Toshihisa Toyoda,Inthiphone Xaiyavong
• Date: 01 March 2016
• Published date: 01 March 2016
• DOI: http://doi.org/10.1111/asej.12085

Currency Substitution in Laos

†

Inthiphone Xaiyavong and Toshihisa Toyoda*

Received 25 May 2013; accepted 26 November 2015

This study investigates the determinants of currency substitution in Laos based on

the function of money demand. The empirical model is estimated using the

autoregressive distributed lag approach to cointegration for the period 1993–2012.

Our estimation results indicate that the interest rate differential is a significant

currency substitution determinant in the Laos economy.Moreover, there is evidence

supporting the existence of a ratchet effect in the currency allocation of deposits,

implying that particularly strong policies should be pursued over an extended period

of time in order to convince depositors to switch back to kip-denominated assets.

Keywords: currency substitution, error-correction model, Laos.

JEL classification codes: F31, E41.

doi: 10.1111/asej.12085

I. Introduction

The feature of currency substitution (CS) in Laos is peculiar and more complicated

than the case of other economies with high CS levels (e.g. Cambodia and some

Latin American countries. The Laos economy is characterized by two main foreign

currencies, the US dollar and the Thai baht, which are widely used in parallel with

the domestic currency, the kip. This means that foreign currencies serve not only as

units of account or as stores of value, but also as media of exchange. Following

Calvo and Végh (1992), we define CS as the phenomenon of use of multiple cur-

rencies as media of exchange as well as units of accounts and stores of value in a

developing country.

The presence of CS has important implications for the conduct of monetary pol-

icy, exchange rate determination, and the stability of demand for money functions.

First, monetary policy becomes less effective because changes in the domestic

money supply or the domestic monetary base have less impact ondomestic expen-

ditures. Second, the volatility of exchange rates is likely to increase. With two or

* Xaiyavong: Monetary Policy Department, Bank of the Lao PDR, Yonnet Road, Chanthabouly

District PO Box 19 Vientiane Capital, Lao PDR. Email: ning.inthiphone@yahoo.com. Toyoda (corre-

sponding author): Professor Emeritus, Kobe University and Hiroshima Shudo University. Postal ad-

dress: 2-3-23-302, Shironoshita, Nada-ku, Kobe, 657-0804, Japan. Email: ttoyoda@port.kobe-u.ac.jp

†

This paper was originally written when the first author was a PhD student at Hiroshima Shudo

University. We are most grateful to Sithanonxay Suvannaphakdy and two anonymous referees for

many detailed comments on this paper.

© 2016 East Asian Economic Association and John Wiley & Sons Australia, Ltd

Asian Economic Journal 2016, Vol.30 No. 1, 67–89 67

bs_bs_banner

more currencies in circulation, the demand for domestic cur rency may become

more sensitive to changes in its expected opportunity costs. Third, the presence

of CS causes instability in the demand for money functions. As the demand for

domestic money becomes less stable, the impacts of monetary policy measures

on domestic expenditures are harder to predict.

In viewof the adverse policy implications of widespread CS, Laos policy-makers

have acknowledged that a strategy for reducing the degree of CS is important in

stabilization and adjustment programs. However, a serious impediment to the

design of such a strategy is a scarcity of empirical studies that provide informative

knowledge on the determinants of the CS. Reliable estimates of the CS determi-

nants can provide implications of CS for the design, execution and effectiveness

of monetary, fiscal and exchange rate policies. The objective of this paper is to

address this issue by examining the determinants of CS in Laos over a two-decade

period within a cointegration and error-correction analytical framework.

Because this study empirically investigates the determinants of CS using quar-

terly data for Laos over the period 1993–2012, it differs from two major studies

that empirically investigate the causes of CS in Laos using time-series analysis.

These studies include de Zamarocksy and Sa (2003) and Ra (2008). De

Zamarocksy and Sa (2003) conduct a causality test between inflation and CS,

using monthly data from February 1995 to September 2001. These authors use

the Granger (1969) methodology to test for the direction of causality between

the two variables. Furthermore, Ra (2008) examines whether the holdings of US

dollar deposits depend on the effect of the expected rate of depreciation in market

exchange rates, using monthly data over the period 1993–2007. Our study differs

from de Zamarocksy and Sa (2003) and Ra (2008) in three important respects.

First, we use the error-correction framework to report both long-run and short-

run impacts. Second, instead of using the Johansen method, we use the bounds

testing approach to cointegration within an autoregressive distributed lag (ARDL)

framework, which was developed by Pesaran and others (Pesaran and Pesaran,

1997; Pesaran and Shin, 1999; Pesaran et al., 2001). Finally, we include the inter-

est rate differential and the ratchet variable in modeling CS in Laos.

While most previous empirical studies of CS have estimated long-run and

short-run impacts using the cointegration and error-correction framework, our

study applies the bounds testing approach to cointegration to examine the dynam-

ics of CS. Estimates using the Engle and Granger (1987) and Johansen (1988)

methods of cointegration are not robust for small sample sizes (Mah, 2000). As

Hakkio and Rush (1991) indicate, expanding the number of observations through

using monthly or quarterly data does not increase the robustness of the

cointegration results, because what matters is the length of the period, rather than

the number of observations. However, a number of Monte Carlo experiments

conducted in Pesaran and Shin (1999) show that the ARDL-based estimation

procedure can be used in small samples to estimate and test hypotheses on the

long-run coefficients when the regressors are I(1) or I(0). For instance, the ARDL

test statistics that are computed using the delta method perform better in small

ASIAN ECONOMIC JOURNAL 68

© 2016 East Asian Economic Association and John Wiley & Sons Australia, Ltd