Download Expectations and Exchange Rate Volatility

THE COST OF CRISES AND LEARNING TO LIVE WITH
EXCHANGE RATE VOLATILITY: EVIDENCE FROM
SURVEY MEASURES OF CONSUMER AND BUSINESS
EXPECTATIONS
Gordon de Brouwer*
Introduction
In many emerging economies there is widespread concern about the possible negative
effects of exchange rate volatility on trade, investment and economic growth.1 There
is, indeed, an enormous empirical and theoretical literature on the possible effects of
exchange rate volatility on an economy.2 Not surprisingly in such a big literature, the
results are ambiguous.
Even if exchange rate volatility is potentially costly, there are two main
arguments for thinking that it will not in fact end up hurting economic activity as
much as anticipated. The first is that those who are concerned about exchange rate
volatility can use financial markets to insure and protect themselves against it.
Exporters and importers, for example, can hedge against future possible exchange rate
movements through forward, future, swap and options markets. These markets can be
substantial. Figure 9.1, for example, shows average daily onshore spot and derivatives
trade in the Australian dollar since 1990. While the spot market has grown, it is much
smaller than the derivatives market (which mostly comprises swaps): there is an
average of about A$50 billion traded in Australian dollar derivatives in Australia each
day. Even more than this is traded offshore.3
Insurance is not free, so there are costs associated with hedging. If the economic
costs of hedging are smaller than the economic costs of volatility, something is gained
by hedging. But there are two important limitations to the claim that households and
firms can hedge their risks. One is that hedging markets may not be well developed in
emerging markets. The BIS (2002) survey estimates of trading in financial markets
show that derivatives trading in foreign exchange in emerging East Asian economies
is relatively low (Table 9.1). Another limitation to the view that financial risk can be
managed by using derivatives is that some sections of business, notably small
business, may not fully use hedging even if it exists. This may be because of the upfront costs involved or because of a lack of business sophistication.4
[FIGURE 9.1 NEAR HERE]
[TABLE 9.1 NEAR HERE]
The second argument for thinking that the costs of exchange rate volatility need
not turn out to be all that high is that even if volatility is initially costly, people
eventually learn to live with it and so it becomes less costly to them over time. In the
context of the debate about exchange rate volatility in East Asia, adaptive learning
implies that the rise in volatility associated with a shift toward more flexible exchange
rate regimes is a transitory adjustment cost.
There is some appeal in this argument. The rational expectations revolution
showed the importance of expectations in explaining economic activity, and
highlighted the notion that people who take account of all available information adjust
their behaviour as the environment around them changes.5 The ever-expanding
literature on the impact of ‘news’ on economic variables, especially financial
variables, is a demonstration of the importance of how people respond to new
information.6 More recently, the literature has focused on adaptive learning
mechanisms in models of expectations. These highlight the notion that ‘rational
expectations’ is an equilibrium concept and that learning and adaptation are part of
the process in reaching the rational expectations equilibrium over time.7
The insight from the learning models can be brought to bear in thinking about
the costs of exchange rate volatility and whether they dissipate as people learn that
exchange rates go up and down over time (even if they are not mean-reverting
processes). One way to examine this is to test whether exchange rate volatility affects
expectations of households and firms – and implicitly, if expectations are important,
their economic behaviour. In this paper, private expectations are measured by
statistical surveys of household and business sentiment. Survey measures are a good,
but imperfect,8 measure of expectations.
The effects of exchange rate volatility on survey measures of household and
business expectations are assessed for a range of East Asian economies. These include
Australia, Japan, Korea, Malaysia, and Singapore. All these countries have survey
2
measures of consumer and/or business sentiment and have (or have had) flexible
exchange rate systems with some variability in the exchange rate.
The methodology is straightforward. The measure of sentiment is regressed not
just on the recent values of exchange rate volatility but also on lags of the dependent
variable and past values of the exchange rate, economic growth, the interest rate and
stock prices. These conditioning variables are included to minimise the risk of bias
from omitted variables.
Time-varying effects of volatility on expectations are shown by looking at subsample effects and plots of recursive estimates of the coefficient on volatility.
Recursive estimates show how the estimated relationship evolves over time as more
observations are added to the sample, so they are especially useful for examining the
impact of particular events (like financial crises) and how the expectations of
households and firms in countries with longstanding flexible exchange rate
arrangements have evolved.
Examining how expectations and the behaviour of consumers and firms have
been affected by exchange rate volatility is a way to answer the question posed by the
learning models about whether the adverse effect of volatility diminishes over time, as
would be the case if people learned to deal with uncertainty.
It also allows a second, more general, question to be addressed: under what
conditions does volatility in the exchange rate adversely affect expectations about the
future? In particular, recursive estimates provide insight into whether there is a
difference between the impact of exchange rate volatility on household and firm
expectations in ‘normal’ times as opposed to extreme events like financial or currency
crises. If exchange rate volatility affects sentiment largely only in crises, then the onus
is on policymakers to create regimes which reduce the risk of crisis (rather than
focusing primarily on regimes which seek to reduce exchange rate volatility).
For the purposes of this analysis, exchange rate volatility is measured by the
period standard deviation of the daily percentage change in the nominal bilateral
US dollar exchange rate, where the period is either a month or a quarter, depending on
the periodicity of the sentiment series.9 The focus is on the volatility of the nominal
bilateral dollar exchange rate because this is the exchange rate which people actually
trade and is the one which is reported in the media.
3
The paper has the following structure. The next section outlines the results from
regressing measures of business sentiment in selected East Asian countries on
exchange rate volatility and a set of conditioning variables. The section after that does
the same for measures of consumer sentiment. The results for both sets are discussed
together in the assessment. A conclusion closes the paper. Descriptions and sources of
the data used in the paper are set out in Appendix 9.1.
Business sentiment and exchange rate volatility
This section explores the relationship between measures of business sentiment and
exchange rate volatility for selected economies in East Asia. Graphs and tables
introduce the data and the results. Figure 9.2 plots the series for Australia, Japan,
Korea, Malaysia and Singapore. Table 9.2 presents the results of simple models of
survey measures of business sentiment for these countries.10 Business sentiment is
regressed in each case on its own lags, recent lags of growth in output (GDP) and
recent lags of exchange rate volatility, the short-term money market interest rate, and
percentage changes in the period-end exchange rate and stock price.
[FIGURE 9.2 NEAR HERE]
[TABLE 9.2 NEAR HERE]
Contemporaneous values of the regressors – notably, exchange rate volatility –
are not included in the equation. This is to avoid the problem of simultaneity.11 The
results are reported for the full sample period.
[FIGURE 9.3 NEAR HERE]
What is of interest is not just the effect of recent exchange rate volatility on
sentiment over the full sample period but also how it has evolved over time. One way
of seeing this is to estimate the regression recursively over time, starting with a small
sample period and extending it one observation at a time. Figure 9.3 plots the
recursive estimate of the regression coefficient of business sentiment on exchange rate
volatility and its two standard error bands, starting with 10 degrees of freedom at the
start of the sample period and ending up with the final estimate shown in Table 9.2.
The data and results have two striking elements.
First, the equations generally explain business sentiment well. There are a
number of plausible regularities. Sentiment tends to be sluggish and persistent (this is
especially the case in Japan and Korea). It generally improves when the economy has
4
been growing and stock prices on the domestic exchange have been rising. Sentiment
tends to fall when interest rates are rising, and, for Japan and Singapore, this is
statistically significant. It also tends to fall when the exchange rate depreciates (as
measured, a rise in the exchange rate), although this is statistically significant only for
Malaysia.
Second, exchange rate volatility negatively affects business sentiment but it is
only statistically significant in the cases of Korean business sentiment and
Singaporean business sentiment. In both cases, the relationship is only statistically
significant from the Asian financial crisis onwards, and in the Singaporean case, it is
not significant at the 5 per cent level (as shown in Figure 9.3). As shown by the
recursive estimates, the negative effect of exchange rate volatility on business
sentiment became bigger during the East Asian financial crisis in all economies
except Malaysia, which fixed its exchange rate to the US dollar in September 1998.
The effects for Singapore depend on the industry: manufacturing firms, for example,
are more sensitive than those in the finance sector to exchange rate volatility.12
The results for the full sample and the recursive regressions point to two
tentative assessments about the impact of exchange rate volatility, to which I return in
the assessment section. In the first place, exchange rate volatility has an
unambiguously negative effect on firms’ expectations about the future, but it is not
generally significant. Furthermore, what really hurts firms’ assessment of their
economic prospects is the variability and uncertainty associated with financial crises.
Crises matter; exchange rate volatility in normal times does not.
Consumer sentiment and exchange rate volatility
This section attempts to model how householders respond to exchange rate volatility.
Figure 9.4 plots the various survey measures of consumer sentiment and exchange
rate volatility for Australia, Japan, Korea and Malaysia.
[FIGURE 9.4]
Table 9.3 presents regression results for the overall sample for each country.
Consumer sentiment is regressed on its own lags, lags of real income or household
consumption growth, lags of the interest rate, exchange rate volatility, and percentage
changes in the period-end exchange rate and domestic stock price. Figure 9.5 plots the
5
recursively estimated regression coefficient of consumer sentiment on exchange rate
volatility, from 10 degrees of freedom to the end of the sample period.
[TABLE 9.3 NEAR HERE]
There are two points to note. First, as for business sentiment, consumer
sentiment can be well explained by a series of economic features. Consumer
sentiment is fairly persistent and it generally improves when demand is growing. It is
negatively associated with domestic short-term interest rates (but only significantly in
the case of Australia) and exchange rate depreciations (but only significantly for
Australia and Malaysia). It is positively associated with domestic stock market
changes (significantly so for Australia and Japan).
Second, consumer sentiment does not generally fall when exchange rate
variability has risen in the recent past, with the notable exception of Malaysia. In this
case, the association was statistically significant during the mid-1990s. The negative
effect of exchange rate volatility on consumer sentiment in Malaysia dissipated
sharply with the introduction of a fixed parity to the US dollar in September 1998. For
Australia and Japan, unlike for business sentiment, the rise in volatility associated
with the East Asian financial crisis does not appear to have had a notable adverse
impact on consumer sentiment.
[FIGURE 9.5 NEAR HERE]
Assessment
Many of the graphs of business and consumer sentiment with exchange rate volatility
in Figures 9.2 and 9.4 suggest that these measures of sentiment react negatively to
exchange rate volatility. It is misleading to read too much into this.
In the first place, bivariate graphical representations may suggest a relationship
which does not exist when tested in a multivariate statistical framework. While the
figures may suggest that movements, especially rises, in exchange rate volatility are
associated with falls in consumer and business confidence, this does not always mean
that there is a statistically significant relationship in a multivariate context with the
right set of conditioning variables. As shown in Figure 9.2, Japanese business
sentiment fell sharply and yen volatility peaked in 1998. This does not mean that the
latter caused the former: this episode is explained by the sharp fall in Japanese stock
prices at the time.13
6
Moreover, there is a risk in over-interpreting the statistical significance of the
volatility variable. While statistical significance may indicate that expectations about
the future are affected by volatility, it may be the case that volatility is merely
capturing some other effect. If the exchange rate is an endogenous price in the
economy, volatility in the exchange rate may reflect some other factor or shock in the
economy, like a domestic policy shock (for example, fiscal, monetary or financial
sector shocks) or a severe external shock (for example, a regional or global crisis). It
may be this factor or shock that affects sentiment and expectations rather than
volatility itself.
Finally, there is also a danger in over-interpreting the results from what is only a
small sample of five countries and, in some cases, relatively small sample periods.
Given these cautions, it is still possible to make some observations about the
results.
First, firms seem to be more sensitive to exchange rate volatility than consumers
are. The coefficient on volatility is negative in all the business sentiment equations but
in only one of the consumer sentiment equations (for Malaysia). The coefficient is
more often statistically significant in the business sentiment equations. This outcome
fits with the well-documented stylised fact that consumption is more persistent than
income and investment.14
Second, exchange rate volatility seems to matter more to sentiment in times of
an exchange rate crisis. As shown in the plots of the recursive coefficient, exchange
rate volatility does not significantly hurt business confidence in Australia, Korea and
Singapore except during periods of extreme uncertainty and instability such as
occurred during the East Asian financial crisis. The damage of exchange rate
volatility on sentiment only seems to arise in extreme events. Exchange rate crises are
costly, more than exchange rate volatility itself. The implication here is that selection
of the exchange rate regime to minimise the risk of crisis is important.
Third, the experience of countries with longstanding flexible currency regimes
suggests that people may learn how to live with exchange rate volatility, such that
volatility generates less uncertainty over time and is less damaging to confidence. In
the Singaporean manufacturing (but not finance) business sentiment equation, for
example, the coefficient on exchange rate volatility is not statistically significant –
7
except for the crisis period – and it diminishes over time. In the Australian and,
particularly, the Japanese consumer sentiment equations, the coefficient on exchange
rate volatility shifts from being negative to positive over time. An interpretation
consistent with this is that people have learned over time that exchange rates can
move up and down by relatively large amounts without causing palpable distress to
the economy. Consequently, the effects on sentiment have dissipated over time.
Fourth, the results for Malaysia are the clearest statement that exchange rate
variability can matter, even outside crisis periods. Oddly, this is most evident for
consumer sentiment than for business sentiment. Not surprisingly, in both cases, the
shift to fixing the ringgit to the US dollar in September 1998 eliminated the effect of
volatility on sentiment.
Conclusion
This paper has assessed the empirical relationship between exchange rate volatility
and survey measures of household and business confidence in Australia, Japan,
Korea, Malaysia, and Singapore.
Caution needs to be used in interpreting this relationship, because the number of
countries and observations is relatively small and because volatility may be a proxy
for the effects of other factors and shocks on sentiment. But some tentative
conclusions can be drawn. Business sentiment is more sensitive than consumer
sentiment to exchange rate volatility. Exchange rate volatility matters much more
when there is a currency crisis than in ‘normal’ times: it seems that currency crises are
much more damaging to confidence than is exchange rate volatility itself. Finally,
there is some evidence that consumers and firms learn to live with exchange rate
uncertainty in flexible rate regimes.
References
BIS (Bank for International Settlements) (2002), Trienniel Central Bank Survey, Foreign Exchange and
Derivatives Market Activity in 2001, Basle.
Blanchard, O.J. and S. Fischer (1996), Lectures on Macroeconomics, MIT Press, Cambridge MA.
Brischetto, A. and G.J. de Brouwer (1999), ‘Householders’ Inflation Expectations’, Reserve Bank of
Australia Research Discussion Paper No. 1999-03.
8
Campbell, J.Y., A.W. Lo and A.C. MacKinlay (1997), The Econometrics of Financial Markets,
Princeton University Press, Princeton.
Engle, R.F., T. Ito and W.L. Lin (1990), ‘Meteor showers or heat waves? Heteroskedastic intra-daily
volatility in the foreign exchange market’, Econometrica, 8(3): 525–542.
Evans, G. and H. Honkapohja (2001), Learning and Expectations in Macroeconomics, Princeton
University Press, Princeton, N.J.
Frenkel, J.A. and M.L. Mussa (1980), ‘The efficiency of foreign exchange markets and measures of
turbulence’, American Economic Review, 70(2): 374–381.
Froot, K. and J. Frankel (1989), ‘Forward discount bias: is it an exchange risk premium’, Quarterly
Journal of Economics, 104(1): 139–61.
Grimes, A., F. Holmes and R. Bowden (2000), ‘An ANZAC Dollar? Currency Union and Business
Development’, Institute of Policy Studies, Victoria University of Wellington, Wellington.
Lucas, R. (1972), ‘Expectations and the neutrality of money’, Journal of Economic Theory, 4, 103–24.
Muth, J.F. (1961), ‘Rational expectations and the theory of price movements’, Econometrica, 29: 315–
35.
Rose, A.K. (2000), ‘One money, one market: the effect of common currencies on trade’, Economic
Policy, 30: 9–45.
Sargent, T. (1973), ‘Rational expectations, the real rate of interest and the natural rate of
unemployment’, Brookings Papers on Economic Activity, 2: 429–72.
9
Figure 9.1
Spot and derivatives turnover in the onshore Australian
dollar market
60
50
30
derivatives
20
10
spot market
0
Se
p90
Se
p91
Se
p92
Se
p93
Se
p94
Se
p95
Se
p96
Se
p97
Se
p98
Se
p99
Se
p00
Se
p01
Se
p02
$A billion
40
Source: Reserve Bank of Australia Bulletin Table F.9
10
Table 9.1
Over-the-counter total foreign exchange derivatives:
average daily turnover, April 2001
US$ milliona
40,852
49,388
534
115,946
3,950
895
3,056
605
69,258
1,669
1,315
Australia
Hong Kong SAR
Indonesia
Japan
Korea
Malaysia
New Zealand
Philippines
Singapore
Taiwan
Thailand
Notes:
a
net of local inter-dealer double-counting.
Source: BIS (2002: 94).
11
Business sentiment and exchange rate volatility
Japan
118
7
105
6
93
5
sentiment (LHS)
80
4
68
3
70
2.5
40
2.0
30
1.5
exchange rate volatility (RHS)
1.0
1
10
0.5
0
0
0.0
2
Ja
n8
Ja 5
n8
Ja 6
n8
Ja 7
n8
Ja 8
n89
Ja
n9
Ja 0
n9
Ja 1
n9
Ja 2
n9
Ja 3
n9
Ja 4
n9
Ja 5
n9
Ja 6
n9
Ja 7
n9
Ja 8
n9
Ja 9
n0
Ja 0
n01
0
-0
ec
02
nJu
D
7
-9
ec
99
nJu
D
4
-9
ec
96
nJu
D
1
-9
ec
93
nJu
D
8
-8
ec
90
nJu
D
87
nJu
Singapore (Finance)
1.1
60
50
1.0
50
0.9
40
0.9
30
0.8
20
0.7
10
0.6
0
0.5
-10
0.4
-20
0.3
-30
0.2
0.8
20
0.7
10
0.6
0
0.5
-10
0.4
-20
0.3
-30
0.2
-40
0.1
-40
0.0
-50
Source: CEIC database
12
1.1
1.0
0.1
exchange rate volatility (RHS)
Ju
nJu 81
nJu 82
nJu 83
nJu 84
nJu 85
nJu 86
nJu 87
nJu 88
nJu 89
nJu 90
nJu 91
nJu 92
nJu 93
nJu 94
nJu 95
nJu 96
nJu 97
nJu 98
nJu 99
nJu 00
n01
exchange rate volatility (RHS)
Ju
nJu 81
nJu 82
nJu 83
nJu 84
nJu 85
nJu 86
nJu 87
nJu 88
nJu 89
nJu 90
nJu 91
nJu 92
nJu 93
nJu 94
nJu 95
nJu 96
nJu 97
nJu 98
nJu 99
nJu 00
n01
-50
net balance
30
sentiment (RHS)
0.0
standard deviation (per cent)
sentiment (RHS)
standard deviation (per cent)
60
40
standard deviation (per cent)
3
0
M
M
ar
-0
ar
-0
7
4
M
ar
-9
1
ar
-9
3.0
50
Singapore (Manufacturing)
net balance
3.5
60
20
exchange rate volatility (RHS)
30
4.0
business sentiment (LHS)
per cent
8
80
index = 100
130
43
M
ar
-7
M
Malaysia
9
standard deviation (per cent)
index
Korea
143
55
8
0.0
3
0.2
-50
0
8
ar
-0
M
6
ar
-9
M
4
ar
-9
M
2
ar
-9
M
ar
-9
0
M
M
ar
-9
8
6
ar
-8
M
4
ar
-8
M
ar
-8
M
ar
-8
M
ar
-8
M
2
0.000
0
-80
0.3
-40
ar
-9
0.175
6
exchange rate
volatility (RHS)
-60
0.5
-30
M
0.350
0.6
-20
ar
-7
-40
0.8
-10
5
0.525
M
-20
0.9
0
2
0.700
ar
-8
0
1.1
10
M
0.875
9
20
1.2
sentiment (LHS)
20
ar
-8
1.050
30
M
40
1.4
exchange rate
volatility (RHS)
M
sentiment (LHS)
1.5
40
ar
-7
1.225
M
60
50
net balance
1.400
standard deviation (per cent)
net balance
Australia
80
ar
-8
Figure 9.2
Table 9.2
Modelling survey measures of business sentiment
Australia
Japan*
Korea*
Malaysia*
Singapore
Singapore
Manufacturing
General
General
General
Manufacturing
Finance
Quarterly
Quarterly
Monthly
Quarterly
Quarterly
Quarterly
1990 Mar qtr
2003 Jun qtr
1974 Sep qtr
2003 Jun qtr
1985 Mar
2003 August
1987 Dec qtr
2003 Jun qtr
1985 Sep qtr
2003 Jun qtr
1985 Sep qtr
2003 Jun qtr
Constant
26.28
[0.20]
–2.07
[0.46]
29.74
[0.00]
11.16
[0.03]
42.65
[0.57]
15.71
[0.02]
Business
sentiment (t-1)
0.47
[0.00]
0.88
[0.00]
0.90
[0.00]
0.81
[0.00]
0.29
[0.03]
0.34
[0.02]
–0.17
[0.02]
–0.42
[0.00]
–0.09
[0.57]
2.14
[0.00]
–0.36
[0.04]
3.86
[0.00]
2.72
[0.04]
1.43
[0.06]
–0.30
[0.01]
2.70
[0.01]
2.76
[0.04]
–19.02
[0.08]
–19.79
[0.24]
–1.19
[0.02]
–1.77
[0.02]
1.20
[0.21]
–2.76
[0.04]
0.10
[0.39]
0.11
[0.57]
Sector
Periodicity
Sample period
Business
sentiment (t-1)
% ∆ growth
(t-1)
6.05
[0.18]
% ∆ growth
(t-2)
Exchange rate
volatility (t-1)
–30.53
[0.17]
–1.02
[0.73]
–4.11
[0.00]
–0.69
[0.59]
% ∆ exchange
rate (t-1)
-0.13
[0.83]
–0.05
[0.63]
–0.10
[0.52]
–0.27
[0.00]
% ∆ exchange
rate (t-2)
Interest rate
(t-1)
–2.09
[0.14]
–0.32
[0.04]
–0.16
[0.28]
∆ interest rate
(t-1)
∆ stock price
(t-1)
–1.08
[0.14]
1.33
[0.00]
0.29
[0.00]
0.18
[0.01]
R-barsquared
0.64
0.90
0.71
0.70
0.56
0.58
Standard
error
19.91
6.00
9.29
4.27
11.48
16.24
1.10 [0.58]
0.62 [0.73]
77.20 [0.00]
0.53 [0.77]
0.85 [0.65]
0.79 [0.67]
Normality
Notes: marginal significance in parentheses; growth is real GDP apart for Korea where it is industrial
production; normality test is the Jarque–Bera test statistic; bold indicates significance at the 10 per cent
level; * indicates Newey–West adjusted standard errors.
Source: author’s calculations.
13
-140
Source: author’s calculations.
14
Jun-03
-120
Jun-02
-80
Jun-01
-100
Jun-00
-60
Jun-99
-40
Jun-98
-20
Jun-97
0
Jun-96
20
Jun-95
40
Jun-94
Singapore Mfg (quarterly)
Jun-93
60
Jun-92
Jun-03
Jun-02
Jun-01
Jun-00
Jun-99
Jun-98
Jun-97
Jun-96
Jun-95
Jun-94
-30
Jun-91
-20
Mar-03
Mar-01
Mar-99
Mar-97
Mar-95
Mar-93
Mar-91
Mar-89
Mar-87
Mar-85
Mar-83
Mar-81
Australia (quarterly)
Jun-90
-10
Jun-93
10
Jun-89
20
Jun-92
Korea (monthly)
Jun-88
-20
Jun-87
-15
-100
Mar-79
-10
-75
Jun-91
-50
Jun-90
-5
Jun-86
0
-25
Mar-77
5
Jun-89
0
Jun-85
10
25
Jun-84
15
50
Mar-75
20
75
Jun-88
30
Mar-73
100
Jun-83
Jun-87
ar
M 80
ar
M 81
ar
M 82
ar
M 83
ar
M 84
ar
M 85
ar
M 86
ar
M 87
ar
M 88
ar
M 89
ar
M 90
ar
M 91
ar
M 92
ar
M 93
ar
M 94
ar
M 95
ar
M 96
ar
M 97
ar
M 98
ar
M 99
ar
M 00
ar
M 01
ar
M 02
ar
-0
3
M
25
Jun-82
Ja
n8
Ja 5
n8
Ja 6
n8
Ja 7
n8
Ja 8
n8
Ja 9
n90
Ja
n9
Ja 1
n9
Ja 2
n93
Ja
n9
Ja 4
n9
Ja 5
n9
Ja 6
n9
Ja 7
n98
Ja
n9
Ja 9
n0
Ja 0
n0
Ja 1
n0
Ja 2
n03
125
Jun-81
Jun-03
Jun-02
Jun-01
Jun-00
Jun-99
Jun-98
Jun-97
Jun-96
Jun-95
Jun-94
Jun-93
Jun-92
Jun-91
Jun-90
Jun-89
Jun-88
Jun-87
Jun-86
Jun-85
Jun-84
Jun-83
Jun-82
Jun-81
Figure 9.3
Recursive estimates: effect of exchange rate volatility on
business sentiment
Japan
40
Malaysia (quarterly)
30
20
0
10
0
-10
-20
-30
200
Singapore Finance (quarterly)
150
100
50
-50
0
-100
-150
-200
Consumer sentiment and exchange rate volatility
Australia
Japan
Korea
3.5
1.5
50
1.0
40
0.5
30
0.0
standard deviation (per cent)
2
ar
-0
2.0
60
1.5
exchange rate volatility
40
1.0
20
0.5
0
0.0
ar
-0
M
02
nJu
0
-0
ec
D
99
nJu
7
-9
ec
D
96
nJu
4
-9
ec
D
93
nJu
1
-9
ec
D
90
nJu
8
-8
ec
M
2.5
80
D
M
100
87
nJu
M
3.0
Source: CEIC database.
15
3.5
consumer sentiment
120
1
60
ar
-0
0
Se
p00
2.0
ar
-9
9
Se
p99
2.5
70
ar
-9
8
Se
p98
80
M
ar
-9
7
Se
p97
3.0
index = 100
4.0
sentiment (LHS)
M
ar
-9
6
Se
p96
4.0
140
4.5
90
Se
p95
index
100
5.0
exchange rate volatility (RHS)
standard deviation (per cent)
110
160
per cent
5.5
120
ar
-0
Malaysia
140
130
0
0.0
6
00
M
nJa
8
98
M
nJa
ar
-9
6
ar
-9
J
-9
an
2
94
M
nJa
4
92
M
nJa
ar
-9
J
0
-9
an
0
88
M
nJa
8
86
ar
-9
J
an
ar
-9
84
M
nJa
6
82
ar
-8
J
an
0.2
30
0
80
0.00
0.3
32
M
exchange rate volatility
(RHS)
60
0.25
ar
-8
70
0.5
34
4
0.50
0.6
36
ar
-8
80
0.8
38
ar
-8
0.75
0.9
40
2
90
1.1
42
M
1.00
1.2
44
ar
-8
100
index
1.25
M
index = 100
110
1.4
exchange rate
volatility (RHS)
sentiment (LHS)
46
M
1.50
sentiment
(LHS)
1.5
48
standard deviation (per cent)
120
nJa
50
1.75
M
130
M
Figure 9.4
Table 9.3
Modelling survey measures of consumer sentiment
Periodicity
Sample period
Constant
Consumer sentiment
(t-1)
Growth (t-1)
Growth (t-2)
Exchange rate volatility
(t-1)
% ∆ exchange rate
(t-1)
Interest rate (t-1)
∆
Australia
Monthly
1984 Jul
2003 Jul
Australia
Quarterly
1984 Jun qtr 2003 Jun qtr
Japan
Quarterly
1982 Sep qtr 2003 Jun qtr
Korea
Quarterly
1995 Dec qtr
2003 Jun qtr
Malaysia*
Quarterly
1988 Jun qtr
2003 Jun qtr
17.12
[0.00]
0.84
[0.00]
0.02
[0.95]
0.50
[0.00]
2.21
[0.11]
–0.60
[0.00]
-0.37
[0.00]
29.5
[0.00]
0.71
[0.00]
2.05
[0.10]
7.62
[0.01]
0.78
[0.00]
0.43
[0.00]
0.50
[0.00]
1.32
[0.22]
–0.01
[0.66]
0.05
[0.55]
22.82
[0.10]
0.76
[0.00]
0.85
[0.35]
17.36
[0.00]
0.87
[0.00}
–0.17
[0.18]
1.40
[0.71]
–0.23
[0.16]
-0.19
[0.59]
–5.43
[0.00]
-0.38
[0.00]
4.28
[0.35]
0.07
[0.69]
–0.55
[0.00]
interest rate (t-1)
% ∆ stock price (t-1)
R-bar-squared
Standard error
Normality
–1.10
[0.34]
0.12
[0.06]
0.87
4.40
1.52 [0.47]
0.03
[0.06]
0.70
6.88
0.13 [0.93]
0.73
1.62
8.77 [0.01]
0.77
7.33
36.59 [0.00]
0.87
5.52
2.24 [0.33]
Notes: marginal significance in parentheses; growth is nominal retail spending for Australia monthly, real
GDP for Australia quarterly but real private consumption expenditure for all other countries; normality
test is the Jarque–Bera test statistic; bold indicates significance at the 10 per cent level; * indicates
Newey–West adjusted standard errors.
Source: author’s calculations.
16
Source: author’s calculations.
17
Jun-03
Jun-02
-10
Jun-01
-8
Jun-00
-50
Jun-99
-40
-6
Jun-98
-4
Jun-97
-2
Jun-96
0
Jun-95
2
Jun-94
Japan (quarterly)
Jun-93
-25
Jun-92
-15
Jun-91
Australia (monthly)
Jun-90
4
ar
M 80
ar
M 81
ar
M 82
ar
M 83
ar
M 84
ar
M 85
ar
M 86
ar
M 87
ar
M 88
ar
M 89
ar
M 90
ar
M 91
ar
M 92
ar
M 93
ar
M 94
ar
M 95
ar
M 96
ar
M 97
ar
M 98
ar
M 99
ar
M 00
ar
M 01
ar
M 02
ar
-0
3
-5
Jun-89
6
M
D
ec
D -79
ec
D -80
ec
D -81
ec
D -82
ec
D -83
ec
D -84
ec
D -85
ec
D -86
ec
D -87
ec
D -88
ec
D -89
ec
D -90
ec
D -91
ec
D -92
ec
D -93
ec
D -94
ec
D -95
ec
D -96
ec
D -97
ec
D -98
ec
D -99
ec
D -00
ec
D -01
ec
-0
2
5
Jun-88
coefficient
15
Jun-87
ar
-7
3
M
ar
-7
5
M
ar
-7
7
M
ar
-7
9
M
ar
-8
1
M
ar
-8
3
M
ar
-8
5
M
ar
-8
7
M
ar
-8
9
M
ar
-9
1
M
ar
-9
3
M
ar
-9
5
M
ar
-9
7
M
ar
-9
9
M
ar
-0
1
M
ar
-0
3
M
Figure 9.5
Recursive estimates: effect of exchange rate volatility on
consumer sentiment
50
Australia (quarterly)
10
0
25
-10
0
-20
-25
-50
30
Malaysia (quarterly)
20
10
-10
0
-20
-30
-60
-70
-80
Appendix 9.1
Descriptions and sources of data
Exchange rates
End-period bilateral US dollar rates, sourced from the IMF’s
International Financial Statistics (IFS).
Exchange rate volatility
Defined as the standard deviation of daily percentage changes in the
dollar bilateral exchange rate. Daily exchange rate data were
obtained from Bloomberg.
Sentiment series
Obtained from the CEIC database for all countries except Australia,
which were obtained from the Reserve Bank of Australia. The
sentiment series are defined as:
Australia
Japan
Korea
Malaysia
Singapore
Consumer sentiment
Melbourne Institute survey of
consumer expectations.
Business sentiment
Australian Chamber of
Commerce and Industry (ACCI)
survey of business conditions
over the next 12 months.
Consumer sentiment
Economic Planning Agency
Consumer Confidence Index,
seasonally adjusted. CEIC code:
JHMA.
Business sentiment
Bank of Japan, Tankan Survey
of All Enterprises over the next
quarter. CEIC code: JONB.
Consumer sentiment
Consumer Spending Plan survey
of expected household income
over the next 12 months. CEIC
Code: KHGH.
Business sentiment
Federation of Korean Industries,
Business Survey Index of future
composite business condition.
CEIC code: KOFAA.
Consumer sentiment
Malaysian Institute of Economic
Research consumer sentiment
index. CEIC code: MOUAD.
Business sentiment
Malaysian Institute of Economic
Research business conditions
index, forecast. CEIC code:
MOUAA.
Business sentiment:
Survey of overall business
conditions over the next six
months for manufacturing and
for finance. CEIC code: SOIAD
and SOIAC respectively.
18
Notes
*
I am grateful for comments from Prema-Chandra Athukorala, Mardi Dungey, and Warwick
McKibbin.
1
The paper by Masahiro Kawai in this volume is a clear expression of this.
2
See Rose (2000) for a new tilt.
3
BIS (2002) data indicate that onshore trade only accounts for about 40 per cent of total
Australian dollar trade.
4
See Grimes, Holmes and Bowden (2000) for a detailed and persuasive case study in New
Zealand.
5
See Muth (1961), Lucas (1972) and Sargent (1973).
6
The literature on the impact of ‘news’ in financial markets is very large; see, for example,
Engle, Ito and Lin (1990) and Campbell, Lo and MacKinlay (1997).
7
See, for example, Evans and Honkapohja (2001).
8
They are imperfect because respondents participate selectively in surveys and they might not be
telling the truth (Froot and Frankel 1989; Brischetto and de Brouwer 1999).
9
This is a standard way of measuring exchange rate volatility (Frenkel and Mussa 1980 and the
paper by Menzie Chinn in this volume). Exchange rate volatility can also be measured in terms
of real exchange rates or deviations from an equilibrium value which may be time varying. The
general assessment in the literature is that these measures tend to be highly correlated with each
other, implying that the proper focus for debate is not the measure so much as the nature of its
relationship with other economic variables.
10
The inference from adjusted Dickey Fuller tests is that sentiment in all cases is stationary.
19
11
Contemporaneous values could be included and then tested for exogeneity but the concern here
is finding sufficiently good instruments which are correlated with exchange rate volatility but
not with shocks to sentiment. As it turns out, contemporaneous values of volatility are only
significant in the regressions on Korean and Malaysian business sentiment. There may also be
concern that exchange rate changes and exchange rate volatility are correlated, which would
give rise to multicollinearity. As it turns out, these variables are largely orthogonal to each other
with the statistical significance of either variable not changing substantially when the other
variable is excluded from the equation. Estimation is in EViews 4.1.
12
A number of alternative measures of business sentiment from Japan’s Tankan survey were also
assessed, such as manufacturing firms and big and small firms. The results for the general
Tankan measure are replicated for alternative sub-samples. For example, business sentiment of
small manufacturing firms is not adversely affected by exchange rate volatility.
13
When the stock price variable is excluded from the regression, the coefficient on exchange rate
variability becomes negative and statistically significant.
14
See Blanchard and Fischer (1996).
20