1. No category

Why are some exchange rates more
volatile than others? Evidence from
middle-income countries
R Havemann and C Kularatne
September 2007: DRAFT FORMAT
Abstract
To better understand why some currencies are more volatile
than others, this paper considers the cross-country determinants
of exchange rate volatility for a set of middle-income countries.
Overall, the paper …nds that higher levels of reserves reduce
volatility, and it is estimated that an appropriate level of reserves
is approximately 4 12 months of imports. Volatility is increased
by increased uncertainty and loose …scal policy. In addition, a
volatile terms of trade spills over into a volatile currency. From a
policy perspective, whilst it is clear that prudent macroeconomic
policy is the best course of action to reduce exchange rate volatility, the in‡uence of external volatility on the exchange rate (over
which the authorities have no control) should not be underestimated.
JEL Codes: F31, C23
1
Introduction
Floating exchange rates are an increasingly common feature of monetary policy regimes, often in combination with an in‡ation-targeting
framework. The bene…ts of a fully ‡oating exchange rate were set out
as early as Friedman (1958). In particular, many countries rely on the
currency to absorb external shocks. Notwithstanding the strong theoretical case for fully ‡oating exchange rates, Calvo and Reinhart (2002)
still …nd a pervasive ‘fear of ‡oating’–many countries are reluctant to
allow their exchange rates to freely ‡oat, due in no small part to the
Strategic Finance & Strategy, Deloitte Consulting; School of Economics, University of Cape Town
1
fact that ‡oating exchange rates can be extremely volatile. In addition,
all real exchange rates, pegged or ‡oating, are characterized by currency
volatility. In the literature, ‘dirty ‡oats’, where the central bank intervenes to manage the value of the currency, are not found to be necessarily
less volatile than ‘free ‡oats’. Indeed, some researchers have found that
central bank intervention increases volatility.
Of the set of middle-income countries, since the end of 2005, the
South African Rand recorded the biggest depreciation (14,8 percent)
against the dollar. One of the factors behind rand depreciation is ongoing concern about the widening current account de…cit (4,2 percent
of GDP in 2005 and 6,1 per cent in the …rst half of 2006). Figure
1 illustrates that, except for the Chilean peso, every economy among
the ten worst-performing currencies against the dollar this year is expected to record de…cits on their current accounts in 2005 (and eight
of the ten worst currency performers were emerging markets, including South Africa). Three of the ten worst-performing currencies could
record double-digit current account de…cits this year: Iceland (13,8 per
cent of GDP), Nicaragua (16,7 per cent) and Guyana (27,9 per cent).
New Zealand, the only developed country among this group apart from
Iceland, could also record a substantial current account de…cit at 8,9
per cent of GDP in 2006. Turkey, which recorded the third biggest
depreciation, is expected to record a current account de…cit of 6,5 per
cent of GDP this year. These …gures imply that there may exist some
association between currency volatility and changes in the current account de…cit/surplus. Furthermore, middle-income emerging market
economies with widening current account de…cits appear to be more
prone to currency volatility. In contrast, Figure 2 shows that three
countries among the ten best-performing currencies this year is expected
to record surpluses on their current accounts in 2006, namely Indonesia,
Brazil and Sweden. The de…cit countries in this group could also record
relatively low or modest de…cit-GDP ratios ranging between 0,6 per cent
of GDP (Haiti) and 8,3 per cent (Romania).
Hausmann (2005) has argued that this volatility may be due to commodity price volatility. To assess this empirically, this paper extends
the econometric framework of Canales Kriljenko and Habermeier (2004)
and Hviding, Nowak and Ricci (2004) to incorporate commodity prices.
It is found that the Hausmann predictions are indeed empirically correct: After controlling for other macroeconomic variables, term of trade
volatility does contribute to exchange rate volatility. The deteriorating
current account balance in some of the emerging-market economies has
been mitigated by net equity portfolio in‡ows. For example, in South
Africa’s case, net non-resident equity portfolio in‡ows averaged US$0,7
2
billion per month in 2005, helping to fund an average monthly current
account de…cit of US$0,8 billion. However, the dependence of South
Africa on non-resident portfolio in‡ows has raised a potential dilemma:
if equity in‡ows suddenly stop or are reversed, the loss of this rand
stability anchor could put downward pressure on the rand and upward
pressure on in‡ation via the exchange rate pass-through to prices. A deteriorating current account balance may increase rand volatility as the
country’s dependence on capital in‡ows (which are relatively mobile)
increases. Fluid capital ‡ows will adversely a¤ect rand volatility. The
South African government should attempt to ensure good governance,
e¢ cient institutions and strive to create an enabling environment in order to in‡uence the risk-adjusted return on investment in order to reduce
the volatility of capital ‡ows and in turn, reduce currency volatility.
The evidence provided labelling the SA rand as the worst performing
currency in the …rst half of 2006 forms the basis of this study to examine
the determinants of the volatility of the rand. Although the SA Rand is
not the most volatile amongst the set of world currencies for the period
1980-2000,1 this is the period (1980-2003) over which this study is conducted. In this period, the rand is the eighteenth most volatile out of a
sample of …fty developed and developing countries. South Africa ranks
near Hungary (17), Mexico (19) and Paraguay (20). These countries are
also in South Africa’s peer group for many other indicators, not least of
which their international credit rating.
[INSERT FIGURE 1 ABOUT HERE]
[INSERT FIGURE 2 ABOUT HERE]
[INSERT FIGURE 3 ABOUT HERE]
This paper proceeds as follows: Section 2 provides a brief overview
of the theoretical and empirical literature on the relationship between
macroeconomic fundamentals and exchange rate volatility; Section 3 discusses the econometric methodology employed in the analysis; Section 4
provides a brief overview of the data; Section 6 discusses the results; and
lastly Section 7 concludes with some policy implications of this study.
2
Empirical Literature
Firstly, it is important to realize that all countries experience real exchange rate volatility, regardless of the exchange rate regime. This is because the real exchange rate measures both internal prices and external
prices (tradables and non-tradables). A country with a …xed exchange
rate and high in‡ation will experience a rapidly depreciating currency.
Indeed, a ‡exible exchange rate may bring about real exchange rate
1
See Figure 3.
3
stability, by reducing the likelihood of speculative attacks against the
currency.
Clark et al (2004) collected data for almost all IMF-member countries, and found that less ‡exible exchange rate regimes do not necessarily guarantee reduced real exchange rate volatility. It could be argued that pegged currencies have a greater likelihood of becoming ‘freely
falling’, as speculators may force the central bank to abandon a peg when
reserves dry up (see, for example, Krugman 1996). The role of exchange
controls is also an area of debate. Canales-Kriljenko and Habermeier
(2004) …nd that prudential limits on banks’foreign exchange positions
may reduce volatility, by reducing speculative position taking. Glick and
Hutchinson (2005) argue that reducing capital ‡ow restrictions may reduce the likelihood of currencies being prone to speculative attacks and
currency crises by reducing foreign exchange market distortions.
After decomposing daily and intra-daily exchange rate volatility into
a continuous and jump components, Beine et al (2006) investigate the
e¤ect of intervention on realized volatility. After controlling for the
impact of macroeconomic announcements on volatility, they …nd that
co-ordinated exchange rate intervention may be a ‘primary’ cause of
discontinuous jumps in the exchange rate, suggesting that intervention
may exacerbate exchange rate volatility, rather than reduce it.
As shown in Table 1, commodity-exporting countries such as Australia and New Zealand also experience amongst the highest levels of
volatility. A similar …nding is made by Hausmann (2005) and in the
data set compiled below. There is a large literature2 on the link between
commodity-exporting countries and the level of the exchange rate, so it
would follow that volatility in commodity prices may also …lter through
to volatility in exchange rates.
[INSERT TABLE 1 ABOUT HERE]
Turning to the recent empirical literature, it is clear that exchangerate volatility is often due to poor macroeconomic fundamentals. Using
a model selection algorithm, Canales-Kriljenko and Habermeier (2004)
found that the following variables were signi…cant determinants of volatility in the nominal exchange rate:
In‡ation;
Real GDP growth;
Fiscal de…cit (% of GDP); and
External trade (% of GDP)
2
See, for example, Clinton (2001), Cashin et al (2002) and Chen and Rogo¤ (2002).
4
Subsequent research by a former South African IMF team (Hviding,
Nowak and Ricci, 2004) has shown that after controlling for macroeconomic conditions, increasing the level of reserves (relative to short-term
debt) also reduces the volatility of the real e¤ective exchange rate. Hviding et al (2004) argue that although theoretically freely-‡oating exchange
rates do not require large reserve holdings, in practice the level of reserves may be an important signal for outside investors, for example, a
low stock of reserves may re‡ect a history of populist monetary policy.
Engel et al. ( 2001) and Mark et al. (2001) …nd evidence that with
the increased e¢ ciency from panel estimation, with the focus on longer
horizons, implies that the macroeconomic models consistently provide
forecasts of exchange rates that are superior to the “no change”forecast
from the random walk model.
3
3.1
Econometric Methodology
Dynamic Heterogenous Panel Model
The natural advantage of using a panel data set and panel estimation
is that the number of data points available becomes su¢ ciently large
to draw meaningful results. Furthermore, using dynamic heterogenous
panel models we can allow for heterogeneity to exist in the short-run dynamics and test for homogeneity in the long-run, between the individual
countries.
Following Pesaran, Shin and Smith (1999), we base our panel analysis
on the unrestricted error correction ARDL(p,q) representation:
4yit =
i yi;t 1 +
0
i xi;t 1 +
p 1
X
j=1
ij 4yi;t j
+
q 1
X
j=0
0
ij 4xi;t j
+
i + "it
(1)
i = 1; 2; ::N , stand for the cross-section units (the countries that
compose the middle-income countries) and t = 1; 2; :::T denote time
periods. Here yt is a scalar dependent variable, xit (k 1) is the vector of
(weakly exogenous) regressors for group i, i represents the …xed e¤ects,
i is a scalar coe¢ cient on the lagged dependent variable, i ’s is the k 1
vector of coe¢ cients on explanatory variables, ij ’s are scalar coe¢ cients
on lagged …rst-di¤erences of dependent variables, and ij ’s are k 1
coe¢ cient vectors on …rst-di¤erence of explanatory variables, and their
lagged values. We assume that the disturbances "it ’s are independently
distributed across i and t, with zero means and variances 2i > 0.
We also make the assumption that i < 0 for all i. This implies that
there exists a long run relationship between yit and xit :
5
yit =
0
i xit + it ,
0
i = i is the
(2)
i = 1; 2; :::; N; t = 1; 2; :::; T
where i =
k 1 vector of the long-run coe¢ cient,
and it ’s are stationary with possible non-zero means (including …xed
e¤ects). Then (1) can be written as
4yit =
+
i i;t 1
p 1
X
ij 4yi;t j
j=1
q 1
X
+
j=0
0
ij 4xi;t j
+
i
(3)
+ "it
where i;t 1 is the error correction term given by (2) and thus i
is the error coe¢ cient measuring the speed of adjustment towards the
long-run equilibrium.
Using the general framework described above,we consider the following three approaches:
1. The Dynamic Fixed E¤ects (DFE) Model imposes the homogeneity
assumption for all of the parameters (across all of the countries in
the middle–country set viz for i = 1; :::; N ) except for the …xed
e¤ects:
= ; i = ; ij = j , j = 1; :::p
1; 2i = 2
ij = j , j = 1; :::q
1
i
(4)
The …xed e¤ects estimates of all the short-run parameters are ob^
tained by pooling and denoted by pooling and denoted by
^
^
^2
^
DF E ,
DF E , j DF E , DF E and DF E . The estimate of the long-run coe¢ cient is then obtained by:
^
DF E
=
0^
@
^
1
DF E A
(5)
DF E
2. The Mean Group (MG) Model estimates allows for heterogeneity
of all parameters and gives the following estimates of short-run
and long-run parameters:
^
MG
=
^
jMG
=
PN
i=1
^
i
N
PN
i=1
N
^
;
^
ij
MG
=
PN
^
i
i=1
N
;
^
, j = 1; ::; p
1;
6
jM G
=
PN
i=1
N
^
ij
, j = 1; ::q (6)
1
^
^
^
^
where i , i , ij and
ually from (1), and
ij
are the OLS estimates obtained individ-
^
MG
1
=N
0^ 1
N
X
@
i=1
^
iA
(7)
i
3. The Pool Mean Group (PMG) model advanced by Pesaran, Shin
and Smith (1999) provides an intermediate case. This estimator
allows the intercepts, the short-run coe¢ cients and error variances
to di¤er freely across groups but the long-run coe¢ cients are constrained to be the same; that is,
i
= , i = 1; 2; :::N
(8)
The common long-run coe¢ cients and the group speci…c shortrun coe¢ cients and the group-speci…c short-run coe¢ cients are
computed by the pooled maximum likelihood (PML) estimation
^
technique. These PML estimators are denoted by
^
i,
^
i,
ij
and
^
ij and
. We then obtain the PMG estimators as follows:
^
P MG
=
^
jP MG
=
^
PN
i=1
^
i
N
PN
i=1
N
^
;
^
ij
P MG
=
PN
i
i=1
N
;
^
, j = 1; ::; p
1;
^
P MG
^
jP M G
=
PN
i=1
N
^
ij
, j = 1; ::q (9)1
=
This depicts the pooling implied by the homogeneity restriction
on the long-run coe¢ cients and the averaging across groups used
to obtain means of the estimated error-correction coe¢ cients and
other short-run parameters. Under long-run slope homogeneity,
the PMG estimators are consistent and e¢ cient.
The Hausman (1978) test3 will be applied to examine the extent of
the panel heterogeneity mainly in terms of the di¤erence between MG
and PMG estimates of long-run coe¢ cients. It should be noted that
the MG estimator provides consistent estimates of the mean of long-run
3
Hereafter we shall refer to the Hausman test as the h-test.
7
coe¢ cients but will be ine¢ cient if slope homogeneity holds. Under longrun slope homogeneity, the PMG estimators are consistent and e¢ cient.
One of the advantages of this estimation approach is that the dynamics are explicitly modelled. Furthermore, we can test whether long run,
homogeneity exists across the countries: i.e. that is that the countries
are su¢ ciently similar to justify arguing that they can be used together
in a panel data set.
3.2
Threshold Autoregressive Estimation (TAR)
The model incorporates Threshold Autoregressive Estimation technique
to investigate the presence of a nonlinear relationship between reserves
and exchange rate volatility.4 We now include an indicator term, which
we use when testing for the existence of a non-linearity. This technique
suggests the estimation of:
yt =
0
+(
11
+
12 I(Pt 1
))Pt
(10)
where yt the log of the coe¢ cient of variation of the real exchange
rate„ P is the policy variable and I(Pt 1
) is an indicator variable.
The indicator variable is created by selecting a potential optimal level of
the policy variable denoted by .
is then subtracted from the original
data series denoted Pt 1 . All values of the new series that are greater
than zero are set equal to one and all values less than zero are set equal
to zero such that I(Pt 1
) is a dummy variable with values of zero
and one. In order to determine what the threshold level might be, we
add the 11 and 12 coe¢ cients. The highest level of reserves in which
any further increases in reserves will either lead to an increase or have
no e¤ect on exchange rate volatility.5
4
The Model
Standard models of exchange rates, based on macroeconomic variables
such as prices, interest rates and output, are thought by many researchers to have failed empirically. However, Mark et al. (2001) provide evidence of exchange rates incorporate news about future macroeconomic fundamentals. We demonstrate that the models might well
be able to account for observed exchange-rate volatility. These models
examine the response of exchange rates to announcements of economic
data.
4
Canning and Pedroni (2004) discusses a separate methodology using the Granger
Represenatation Theorem to test if infrastructure capital stock has a positive or
negative e¤ect on per capita GDP.
5
See Potter (1995) and Koop, Pesaran and Potter (1996).
8
From Engel et al. ( 2001), suppose the exchange rate is determined
by:
Coef f (st ) = xt +
t
+ E (st+1 j
t)
+ "t
(11)
where Coef f (st ) denotes the coe¢ cient of variation of the nominal exchange rate, xt a macroeconomic fundamental, a measure of
risk/uncertainty and E (st+1 j t ) expected nominal exchange rate based
on the information set t . Now suppose we observe only some subset of
the fundamentals that drive the exchange rate, xt .
The above implies that the current exchange rate is a¤ected by both
observable and unobservable variables a¤ect the volatility of the currency.
5
Data
The data set consists of a panel of 11 macroeconomic indicators for 19
countries between 1981 and 2003.
There are a number of di¤erent approaches to measuring exchange
rate volatility and as a result, there is no generally accepted measure.
Most countries do not publish monthly real exchange series, so this paper
creates a real exchange rate series in the standard way, by de‡ating the
nominal exchange rate relative to the United States using that country’s
consumer price index relative to the US consumer price index. However,
the disadvantage is that this is not a trade-weighted measure, although
for a signi…cant proportion of countries in the sample most trade is
denominated in US dollars, suggesting this not a particularly important
problem. The coe¢ cient of variation6 of the currency was calculated
based on this monthly real exchange rate for each country in the sample.
A number of interesting observations emerge from the data. In terms
of volatility against the US dollar, South Africa is ranked with approximately the same level of volatility as Paraguay, Mexico and Hungary.
In this particular sample, Brazil and Turkey are the two most volatile
currencies, whilst Canada is the least volatile.
Figure 4 depicts SA rand volatility from 1980 to 2003. Rand volatility tends to rise in periods of uncertainty. For example, in 1985 and
1994 (Rubicon Speech and …rst democratic elections, respectively), rand
volatility rose.
[INSERT FIGURE 4 ABOUT HERE]
The macroeconomic fundamentals investigated for having an e¤ect on
exchange rate volatility are:import cover, budget surplus-to-GDP ratio,
6
The coe¢ cient of variation is the standard deviation divided by the mean and
is a more appropriate measure of volatility if data series have signi…cantly di¤erent
means.
9
debt-to-GDP ratio, GDP growth rate, a measure of openness of the
economy,7 a measure of uncertainty,8 oil price volatility, change in the
current account and dollar-euro volatility.
6
Estimation Results
During the process of estimation, two groups of variables were found to
have a signi…cant in‡uence on real exchange rate volatility. The …rst
group is the country’s macroeconomic fundamentals, i.e. variables that
carry speci…c information about the country itself. The second group of
variables – commodity price volatility – captures the e¤ect of volatility
in international markets.
[INSERT TABLE 2 ABOUT HERE]
6.1
6.1.1
Macroeconomic variables (‘fundamentals’)
Reserves
Import cover, measured in number of months, is found to be a signi…cant
determinant of exchange rate volatility. It is estimated that an increase
in foreign reserves by 1 month reduces real exchange rate volatility by 4
per cent. This result is robust across speci…cations.
Hviding et al (2004) argue that this result may be non-linear, with
‘decreasing returns to reserves’. Testing for this, we …nd that the optimal
level of reserves South Africa should hold (in terms of import cover) is 4and-a-half months. South Africa currently has approximately 3 months
of reserves. Thus increasing the amount of reserves further should aid in
reducing rand volatility. Furthermore, Hviding et al (2004) argue that
although theoretically freely-‡oating exchange rates do not require large
reserve holdings, in practice the level of reserves may be an important
signal for outside investors, for example, a low stock of reserves may
re‡ect a history of populist monetary policy.
6.1.2
Fiscal policy
Two di¤erent variables were used to measure the …scal policy stance - the
budget balance-to-GDP ratio and debt-to-GDP ratio. A better budget
balance-to-GDP ratio (i.e. a narrowing of the budget de…cit or increase
in the budget surplus) reduces real exchange rate volatility. Referring
to regression (1) (in Table 2) it can be seen that a one percentage point
improvement in this ratio leads to a 0.8 percent decrease in real exchange
rate volatility. Similarly a higher debt-to-GDP ratio is associated with
7
Openness is measured as the ratio of the sum of exports and imports-to-GDP.
The measure of uncertainty used in this model is the di¤erence between the real
long-term interest rate in the US and the country in question squared.
8
10
higher real exchange rate volatility. As indicated in regression (2) a
one percentage point increase in the debt-to-GDP ratio leads to a 0.27
percent rise in exchange rate volatility.
Both variables are signi…cant at the 5 per cent level. Note that the
number of countries in the sample falls from 25 to 19 if the debt-to-GDP
ratio is used rather than the …scal balance variable. This is due to a lack
of a consistent debt data set across countries.
6.1.3
GDP growth
This is found to be only weakly signi…cant and not robust to di¤erent
speci…cations. One regression that did include this variable is reported
– regression (4). It suggests that a one percentage point increase in
growth leads to a 2.7 percent decrease in real exchange rate volatility.
However, it cannot be overemphasized that this variable is problematic
in this context. It is not robust across di¤erent speci…cations. There is
also an inherent endogeneity problem with it not being theoretically or
empirically clear if exchange rate volatility reduces growth or if growth
reduces exchange rate volatility (for example, a strongly growing country
may be at less risk of a currency crisis as it able to attract su¢ cient
capital in‡ows).
6.2
Terms of trade shocks
Given the small and open nature of most of the middle-income emerging
markets in the data set, a priori there is reason to believe that the terms
of trade may in‡uence the exchange rate. By extension, terms of trade
volatility should in‡uence exchange rate volatility. To the knowledge of
the authors this has yet to be incorporated in the empirical literature,
although Hausmann (2005) provides a strong theoretical case for how
terms of trade volatility may in‡uence the exchange rate.
Intuitively, under a ‡exible exchange rate regime, countries experiencing volatile terms of trade will also experience volatile exchange
rates; whereas under …xed regimes, the consequences of a sharp increase
in commodity prices (such as the oil price) will be re‡ected in higher
in‡ation. As a result, in‡ation will be volatile and hence the underlying
real exchange rate will become volatile too.
Due to data constraints, it was not possible to construct a direct
measure of the term of trade volatility as this data is not available for
all countries at a high frequency and as the terms of trade measure is
highly correlated with the exchange rate. Second best is thus to proxy
this volatility. This paper chooses oil price volatility as the measure best
suited to proxy terms of trade volatility. This is because the oil price is
an important part of either the import or export basket for almost all
11
the countries in the sample.
6.2.1
Oil price volatility
It is found that a one percent increase in the volatility of the oil price
leads to a 0.06 to 0.07 percent increase in the volatility of the exchange
rate depending on the speci…cation. Table 2 reports two of the regressions, one with the budget-balance measure and one with the GDP
growth measure.9
[INSERT FIGURE 5 ABOUT HERE]
6.2.2
Current account balance
The research conducted found that the level of the current account balance did have a statistically signi…cant e¤ect on rand volatility. A one
percent improvement in the current account will lead to 4 per cent decrease in rand volatility. In addition, improvements in the current account reduce rand volatility at an increasing rate, i.e, the more the current account improves, there are rising gains in terms of reduced rand
volatility
Thus continuous improvements in South Africa’s current account balance, continuously reduces rand-dollar volatility.
6.3
World volatility
There is a strong argument that there are time-speci…c factors at play
when considering exchange rate volatility: i.e. one can identify periods
such as the Asian crisis for example. Although the e¤ects of the sudden devaluation of the Thai bhat during the 1998/9 Asian crisis were
initially only felt in south-east Asia, …nancial contagion quickly spread
throughout global …nancial markets. This is a good example of a case
where the poor fundamentals of one country (in this case Thailand) led
to exchange rate volatility in other (emerging) countries.
There are a number of approaches that may be used to capture this
type of e¤ect. As indicated in Figure 1, the measure of terms of trade
volatility may also be a good indicator of world volatility.
An alternative measure, the volatility in the dollar-euro exchange
rate, is also found to be signi…cant in some speci…cations (see regression
4 in Table 2) for an example. However, in most speci…cations it is found
to be not statistically signi…cant. A one percent increase in dollar-euro
volatility leads to a 2.31 percent increase in a country’s real exchange
rate volatility.
9
Consider regressions (1), (3), (6) and (8) in Table 2.
12
7
Conclusion & Policy Recommendations
The literature and the results in this study suggest that prudent macroeconomic policy (a combination of low in‡ation, health budget balance
and reserves accumulation) lowers exchange rate volatility. Particularly,
results in other studies underscore the role of reserves in reducing volatility, even if this is merely as a signal to market participants.
This study extends the literature to consider the role of the terms
of trade. It is found that terms of trade volatility (proxied by oil price
volatility) increases exchange rate volatility for the countries in the sample. This is not surprising as many of the countries are either signi…cant
importers or exporters of oil.
The relationship between GDP growth and exchange rate volatility appears complex and raises a number interesting puzzles: Does GDP
growth lead to lower exchange rate volatility as capital in‡ows are strong
and stable? Or, does lower exchange rate volatility lead to higher GDP
growth? Or, controversially, does a volatile exchange rate bu¤er the
economy from external shocks? There is a large literature on the relationship between trade and exchange rate volatility (see, for example,
Clark 2004). However, in general the results are mixed and it is not clear
what the nature of this relationship is. Clearly more research is required
regarding the impact of exchange rate volatility on GDP growth in the
South African case.
The policy recommendations from this study are as follows:
1. Prudent …scal management will not only have a direct impact on
exchange rate volatility via the impact on the budget de…cit and
government-debt ratio, but also indirectly via the impact on the
current account;
2. Current account balance is of far greater importance than reserve
levels as far as exchange rate volatility is concerned;
3. Even though reserves has been rising as the current account balance has weakened, the bene…ts from increased reserves will not
make up for the increased uncertainty (volatility) emanating from
the worsening current account; and
4. Government needs to continue to engage with the public by being
even more transparent and sincere in its policy objectives and goals
in order to reduce uncertainty.
This study concentrated on the cross-country determinants of exchange rate. Future research may need to concentrate on country-speci…c
13
models. In particular, the relatively concentrated market structure raises
questions about whether or not participants have su¢ cient power to
‘move’the market, or whether large transactions create temporary exchange rate misalignment, which then is corrected after a period of relative volatility.
14
8
References
References
[1] Beine M., J. Lahaye, S. Laurent, C. J. Neely and F. C. Palm. (2006).
‘Central Bank Intervention and Exchange Rate Volatility’. Federal
Reserve Bank of St Louis Working Paper 2006-031A.
[2] Calvo, G. A. and C. M. Reinhart. (2002). ‘Fear Of Floating,’Quarterly Journal of Economics, v107 (May), pp379-408.
[3] Canales Kriljenko, J. I. , K. F. Habermeier. (2004). ‘Structural Factors A¤ecting Exchange Rate Volatility: A Cross-Section Study’.
IMF Working Paper No. 04/147
[4] Cashin, P. L. Céspedes and R. Sahay. (2001). ‘Keynes, Cocoa and
Copper: In Search of Commodity Currencies’. IMF Working Paper
WP/02/223. Available online http://www.imf.org
[5] Chen, Y. and K. Rogo¤. (2002). ‘Commodity Currencies and Empirical Exchange Rate Puzzles’. IMF Working Paper WP/02/27.
Washington D.C.: Own publication.
[6] Clark, P. B., N. T. Tamirisa, S-J Wei, A. M. Sadikov L. Zeng.
(2004). ‘A New Look at Exchange Rate Volatility and Trade Flows’.
IMF Occasional Paper No. 235
[7] Clinton, K. (2001). ‘On Commodity-Sensitive Currencies and In‡ation Targeting’. Bank of Canada. Working Paper 2001-3.
[8] Glick, R. and M. Hutchison. (2005). ‘Capital controls and exchange
rate instability in developing economies’. Journal of International
Money and Finance, Elsevier, vol. 24(3), pages 387-412.
[9] Engel, C., Mark, N., and West, K. (2007). ‘Exchange rate models
are not as bad as you think’. NBER Working Paper No. 13318.
[10] Friedman, M. (1958). ‘The Supply of Money and Changes in Prices
and Output’ in Relationship of Prices to Economic Stability and
Growth.
[11] Hausmann, R. (2005). ‘Ensuring a Competitive and Stable Real
Exchange Rate in South Africa’. Mimeo: Preliminary research from
South African Treasury Economic Research Programme.
[12] Hviding, K., Nowak, M.and Ricci, L. A. (2004). ‘Can Higher Reserves Help Reduce Exchange Rate Volatility?’ IMF Working Paper
WP/04/189
[13] Krugman, P. (1996). ‘Are Currency Crises Self-Ful…lling?’, in Ben
S. Bernanke and Julio J. Rotemberg eds. NBER Macroeconomics
Annual 1996, MIT Press, Cambridge Mass. and London
[14] Mark, Nelson A., Sul, D. (2001). ‘Nominal Exchange Rates and
Monetary Fundamentals: Evidence from a Small Post-Bretton
Woods Sample’, Journal of International Economics 53, 29-52.
15
A
Appendix
Ten-worst performing currencies v/s US$
SA Rand
Iceland Kro na
Turkish Lira
M auritius Rupee
Nicaragua Co rdo ba
Guyana Do llar
Chilean P eso
Co lo mbian P eso
Co sta Rican Co lo n
New Zealand Do llar
Current acco unt, % o f GDP , 2006
% change since 31Dec 2005
-28.0 -24.0 -20.0 -16.0 -12.0 -8.0
Figure 1: T en worst
Indonesian Rupiah
Brazil Real
Slovakian Koruna
Swedish Krona
UK Pound
Haiti Gourde
Czech Koruna
Thai Baht
Paraguay Guarani
New Romanian leu
Figure 2:T en best
0.0
4.0
perf orming currencies in the f irst
Ten best performing currencies v/s US$
-12.0
-4.0
-7.0
-2.0
3.0
Current acco unt, %
o f GDP , 2006
% change since 31
Dec 2005
8.0
13.0
perf orming currencies in f irst
16
half of 2006
half of 2006
Canada
Malaysia
Australia
United_Kin
Bolivia
New_Zeala
Norway
Sweden
Germany
Italy
France
Austria
Netherlands
Slovakia
Sri_Lanka
Belgium
Iceland
Denmark
Thailand
Korea
Portugal
Switzerland
Finland
Spain
Japan
Gabon
Israel
Greece
Chile
Pakistan
India
Paraguay
Mexico
SOUTH_AF
Hungary
Argentina
Egypt
Slovenia
Columbia
Indonesia
Algeria
Poland
Kenya
Ghana
Uruguay
Venezuela
Ecuador
Peru
Bulgaria
Turkey
Brazil
0
0.1
0.2
0.3
0.4
0.5
0.6
Figure 3.Average exchange rate volatility against U S dollar; 1980
2000:
Most and least volatile exchange rates, 1970 to 2002
Five most volatile currencies
Advanced
Japan, Australia, Israel, New Zealand, United Kingdom
Emerging
Argentina, Uruguay, Turkey, Chile, Indonesia
Developing
Congo (Dem Rep of), Sudan, Angola, Bolivia, Ghana
Five least volatile currencies
Advanced
Austria, BenLux, Canada, Netherlands, Denmark
Emerging
Panama, Singapore, Malaysia, Venezuela, Mexico
Developing
Martinique, French Guiana, Réunion, Netherlands Antiles, Bahamas
Source: Adapted from Clark et al (2004)
Table 1. M ost and least volatile exchange rates (1970
17
2002)
Ra nd-D ollar ex chan ge ra te
0 .18
12
Re al Ra nd /Do llar vo la tility
Re al Ra nd /Do llar Exc h ang e rate
0 .16
Ra nd
colla pse
Ru bicon spe ec h
an d sa nc tion s
10
0 .14
Arg en tin ia n/ Me x ic a
n/R ussian crises
0 .12
0.1
8
A s ia n c ri s is
6
0 .08
0 .06
4
P re-1 99 4
e le ctio n jitters
Ma nd ela' s
relea se
0 .04
2
0 .02
0
0
19 81 1 98 2 1 98 3 19 84 19 85 1 986 1 98 7 1 98 8 198 9 19 90 19 91 1 99 2 1 99 3 1 99 4 19 95 19 96 1 997 1 99 8 1 99 9 200 0 20 01 20 02 2 00 3
Y e ar
Figure 4:Rand
Dollar exchange rate f rom 1980
0.35
Gulf
war
Oil crisis
2003:
Asian
crisis
Coefficient of variation (annual)
0.30
0.25
0.20
0.15
0.10
0.05
02
00
Gold price volatility
Figure 5:Oil and gold price volatility (1970
18
20
98
20
96
19
94
19
92
Oil price volatility
19
90
19
88
19
86
19
84
19
82
19
80
Volatile periods
19
78
19
76
19
74
19
72
19
19
19
70
0.00
2002)
Variables
Macroecon
omic
variables
Import
cover
(weeks)
mpco
Budget
surplus to
GDP ratio
dfgd
Log debt-to
GDP ratio
ldgp
GDP
growth
dgdp
Openness
Log
Uncertainty
lunc
Terms of
trade
volatility
Log Oil
volatility
loiv
Change in
Balance of
Payments
(current
account)
World
volatility
Dollar/euro
volatility
Φ
Joint
Hausman
test (pvalue)
Regress
ion
(1)
Reserve
s,
Budget
surplus
-toGDP,
uncertai
nty, oil
volatilit
y
N = 25
T = 15
Regress
ion (2)
Regress
ion
(3)
Reserve
s, GDP
growth,
uncertai
nty, oil
volatilit
y
Regress
ion (4)
Regress
ion (5)
Regress
ion (6)
Regress
ion (7)
Regress
ion (8)
Regress
ion (9)
Reserve
s, GDP
growth,
uncertai
nty
dollar/e
uro
volatilit
y
Budget
surplus
-toGDP,
opennes
s,
uncertai
nty
Openne
ss,
uncertai
nty, oil
volatilit
y
Change
in the
current
account,
import
cover,
uncertai
nty
Inflatio
n rate,
change
in the
inflatio
n rate,
oil
volatilit
y
Debtto-GDP,
inflatio
n rate,
change
in
inflatio
n rate
N = 19
T = 22
N = 19
T = 21
N = 19
T = 20
N = 19
T = 21
N = 19
T = 23
N=19
T=23
N=19
T=22
N=19
T=22
0.043**
(-2.98)
0.044**
(-3.09)
0.037**
(-4.45)
0.044**
(-3.64)
n/a
n/a
0.022**
(-2.78)
n/a
n/a
-0.008*
(-1.89)
n/a
n/a
n/a
0.013**
(-2.36)
n/a
n/a
n/a
n/a
n/a
0.279**
(2.90)
n/a
n/a
n/a
n/a
n/a
n/a
0.398**
(4.04)
n/a
n/a
-0.574
(-1.50)
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
2.378**
(-4.33)
n/a
-0.221*
(-1.72)
n/a
n/a
n/a
0.103**
(9.65)
0.100**
(9.10)
0.117**
(16.20)
0.132**
(9.25)
0.105**
(8.392)
0.122**
(-2.14)
0.096**
(9.29)
0.100**
(8.97)
n/a
n/a
0.062**
(2.44)
n/a
0.074**
(3.94)
n/a
n/a
0.064**
(2.79)
n/a
0.060**
(2.20)
n/a
n/a
n/a
n/a
n/a
n/a
0.068**
(-3.95)
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
0.858**
(-13.27)
1.80
(0.77)
0.842**
(-12.46)
0.53
(0.91)
0.915**
(-7.24)
7.10
(0.13)
2.309**
(2.82)
-0.827
(-8.85)
0.908**
(-17.90)
4.08
(0.25)
0.985**
(-10.56)
4.14
(0.25)
0.913**
(-7.55)
6.15
(0.10)
-0.86**
(12.90)
-0.79**
(-12.87)
5.99
(0.11)
5.35
(0.15)
Reserve
s, Debtto-GDP,
uncertai
nty
5.39
(0.25)
Table 2. Estimation results
19
Number of Rand cents above the
average2
Macroeconomic
variables
Current values
Scenario
Difference
First half of 2006
Scenario result
Difference3
Import cover
(months)
3*
2
1 month
29 cents
30 cents
1 cent
Budget surplus to
GDP ra tio
(%)
-0.59*
-1.59
1 percenta ge
point
29 cents
49 cents
20 cents
Debt-to GDP
ratio (%)
35.8*
36.8
1 percenta ge
point
29 cents
59 cents
30 cents
29 cents
31 cents
2 cents
Uncertainty
Rise by 50 basis points
Terms of trade
volatility
Oil volatility
Openness (%)
Current account
balance
Standard deviation rises by US$ 1**
29 cents
30 cents
1 cent
56
55
1 percenta ge
point
29 cents
30 cents
1 cent
-4.10*
-5.1
1 percenta ge
point
29 cents
59 cents
30 cents
29 cents
34 cents
5 cents
World volatility
Dollar/Euro
volatility
Standard deviation rises by 1 US cent **
Table 3. Scenarios
20