Table 4 shows the GARCH(1,1) estimation results from regressing changes in the JPY/USD exchange rate on intervention of which the market is aware (INT^RUMOR), intervention of which the market is unaware (INT^NoRUMOR), and the indicator variable capturing rumors of intervention on days when no intervention occurs (RUMOR^NoINT).

The models also include as explanatory variables the two indicator variables capturing intervention statements (PSTAT and NSTAT) as well as the 11 control variables containing various Japanese and US macro surprises.

The first column of Table 4 displays the results pertaining to the full sample (January 1999 to March 2004).

Both INT^RUMOR and INT^NoRUMOR are highly significant (at the 99 percent level) and of the correct (positive) sign, thereby providing evidence that actual intervention (i.e. JPY sales), whether the market is aware of the intervention or not, is, on average, associated with an exchange rate movement in the intended direction (i.e. JPY depreciation).

By contrast, RUMOR^NoINT is highly insignificant, implying that, on average, a rumor of intervention is not in itself sufficient to elicit a detectable exchange rate movement.

As discussed earlier, evidence of effectiveness of intervention of which the market is aware is consistent with a portfolio-balance transmission channel as well as an information signaling transmission channel.

Evidence of effectiveness of intervention of which the market is unaware is attributed to the former channel, while evidence of effectiveness of rumors of intervention on days when no intervention occurs is attributed to the latter channel.

The significance and sign of both INT^RUMOR and INT^NoRUMOR suggest that intervention works through the portfolio-balance channel yet, at the same time, it is not possible to rule out that intervention may also work through the signaling channel.

The insignificance of RUMOR^NoINT shows that no evidence of an information signaling channel is found.

Taken together, a cautious characterization of the full sample baseline results is that the importance of the portfolio-balance channel is confirmed while no evidence of an information signaling channel is found.

Turning to the possibility that official statements (“oral intervention”) impact the exchange rate, the highly insignificant PSTAT and NSTAT variables clearly suggest that this is not the case. This is an interesting and, in light of existing studies such as Beine, Janssen and Lecourt (2004) and Fratzscher (2004, 2005) who find “central bank communication” to influence exchange rate markets, at a first glance a surprising result.

However, considering the time period under study, the rejection of an influence of either “positive” or “negative” statements is perhaps less surprising.

During the period under study, all interventions are unilateral, all interventions are carried out in the same direction (i.e. JPY sales), and the intervention frequency is remarkably high, especially during the latter part of the sample period.

Therefore, it seems plausible that during the period under study there is very little uncertainty in the market about the desire of the intervening authority to depreciate the JPY and, accordingly, “positive” statements repeating or confirming this desire do not contain a sufficient degree of news to affect the market.

Similarly, the insignificance of the “negative” statements may be due to the fact that they appear at odds with the actual intervention operations and, therefore, are disregarded by the market.20

20 Beine, Janssen and Lecourt (2004) analyze the impact of statements over the 1991 to 2003 period and Fratzscher (2004, 2005) analyzes the impact of statements over the 1990 to 2003 period.

Their samples contain unilateral as well as coordinated interventions, intervention in opposite directions, and substantial variation in intervention frequencies.

Both find that official statements impact the JPY/USD exchange rate. Fatum and Hutchison (2002) and, subsequently, Jansen and de Haan (2005) and others find some evidence that official statements influence the EUR/USD exchange rate.

The second and third columns of Table 4 report the results pertaining to the truncated samples.
As the columns show, the Sub-Sample 1 and Sub-Sample 2 findings regarding the effects of intervention, rumor, and statement variables are identical to the previously described full sample.

Overall, the results are robust across the full sample and Sub-Samples 1 and 2.
The fourth column of Table 4, however, reveals drastically different results.
For SubSample 3, none of the intervention variables are significant and, furthermore, the standard F-test rejects the model specification altogether.

In other words, while it appears that intervention of which the market is aware as well as intervention of which the market is unaware both influence the JPY/USD exchange rate over the 1999 to 2003 period, neither categories of intervention have any impact whatsoever on the JPY/USD during the first quarter of 2004.21

As before, rumors of intervention on days with no intervention as well as intervention statements are insignificant.22

21:Although the empirical method of Fatum and Hutchison (2006b), who introduce the matching methodology to the intervention literature, is different from the time-series analysis of this study, they also find a complete absence of effectiveness of official Japanese intervention during the first quarter of 2004.

They conjecture that the lack of effectiveness during the first quarter of 2004 is related to the unusually high intervention frequency of this period.

22:Sub-Sample 3 consists of only 64 observations and, therefore, results pertaining to this particular sample are interpreted with caution.