(3) ∆s,t+k = a + b1 ~e + b2~U + CZt+k + εt, k=0,...60.
where ∆st+k is the first-difference in the log of the spot exchange rate k-periods ahead and Zt+k contains
the control variables k-periods ahead. Since our analysis uses an event study approach instead of time-
series techniques, testing for delayed effects by lagging the exogenous monetary policy variable is not
meaningful. Therefore, leads of the independent variable and the associated news control variables rather
than lags of the monetary policy variables are used for capturing any delayed exchange rate responses to
the monetary policy changes.
We estimate the regression models described by equation 3) using standard OLS with Newey-
West covariances (see Newey and West, 1987) in order to take into account the possibility of
autocorrelation in the control variables.
The estimations cover up to 60 leads of the independent variable (12 business weeks, ensuring
that we capture any delayed effects in-between FOMC dates). For the sake of exposition, we only display
a summary of each of the estimations using 0 through 15 leads (three business weeks), respectively, for
each of the three exchange rate variables.
A quick absorption process would be consistent with monetary policy surprises being
systematically related to same-day changes in exchange rates (i.e. b1 should be significant only when
k=0) while the “current” monetary policy surprises should be orthogonal to “future” exchange rates (i.e.
b1 for k = 1,...,60 should all be insignificant).
3.2 The DEM/USD Exchange Rate
The first two columns of Table 2 show the estimation results from regressing same-day changes in the
DEM/USD exchange rate on the expected and the unexpected components of the Fed funds target rate
changes, as described in equation (1). The first column shows the results when including all the news
control variables while the second column shows the results when including only the significant news
10