13
In the case of Emerging Asia, both procedures detect volatility shifts during 1997-
1998, but none in 1999-2002. In particular, the wavelet variance test finds breakpoints over
1997-1998 at the finest scales (d1, d2, and d3) but not at the coarser ones (d4, d5, and d6).
For North America, the ICSS fails to detect breakpoints between 1999 and 2001, whereas
wavelet analysis finds some evidence against the null hypothesis of variance homogeneity
in that period—in, particular at scales 1 and 6. For 1999-2002, the evidence against the null
hypothesis is mixed, according to wavelet analysis: breakpoints are detected at scale 1 at all
significance levels, and at scales 2 and 3 at the 10 percent and 5 percent of significance
level, respectively.
Finally, for Europe, for 1997-1998, both tests find evidence against the null
hypothesis. However, unlike the ICSS algorithm, the wavelet variance test does not find
any breakpoints in 2000. Regarding 2001, the latter test finds shifts only at scale 1 at all
significance levels, and at the 10 percent level at scale 2. No breakpoints are found at
higher scales of the data. When taking 2001-2002, the null hypothesis of homogeneity is
rejected, at all significance levels, at scales 1, 2, and 3. This suggests that the extra year of
data (2002) adds some variance instability, as detected by the ICSS algorithm.
[Table 2]
Regarding interest rates, wavelet analysis finds breakpoints in both periods for the
30-day, 90-day, and 180-day interest rates, but none for the 60-day interest rate in the first
period. In general, most violations of variance homogeneity are observed in the second
period, in which the change in monetary policy occurred. For instance, variance
homogeneity is rejected at all significance levels at the first four scales of the 30-day
interest rate, and at the first three scales of the 60-day interest rate. Overall, the periods of
breakpoints coincide with those detected by the ICSS algorithm. The exception is the 30-