Clearly, if the Fed had been following a fixed or managed float exchange rate policy from 1989
onwards, US monetary policy changes could have been systematically affected by same-day exchange
rate changes. However, as pointed out by Eichenbaum and Evans (1995) and Kim and Roubini (2000),
the period under study is characterized as a floating exchange rate regime, thus there is no reason to
believe that the Fed changes US monetary policy in response to same-day or short-term exchange rate
movements. Based on this institutional factor, therefore, simultaneous equation bias (endogeneity) should
not be present in our regression models.10
In order to assess the importance of disentangling the surprise component from the actual
monetary policy change we estimate a regression model of the exchange rate response to the monetary
policy change without using the Kuttner-decomposition:
(2) ∆st =α+β1(r~ta)+CZt +εt
where r~ta (= r~te+ r~tu ) is the actual monetary policy change in percentage points. Consistent with Cook
and Hahn (1989) and Kuttner (2001) we estimate the regression models described by equations 1) and 2)
using standard OLS estimation techniques.
In order to test whether exchange rate markets absorb monetary policy news quickly or whether
the absorption process stretches over or takes place after several days we estimate regression models
characterized by the following equation:
of the surprise element of central bank intervention or foreign monetary policy changes. Therefore, we address the
issue of interventions and foreign monetary policy changes in the robustness section while our baseline models
described in this section incorporate only the control variables that are based on expectations and capture surprises.
10 In the context of a time-series analysis of exchange rate responses to day-to-day changes in monetary policy
expectations in-between actual monetary policy changes, Fatum and Scholnick (forthcoming) formally test for
simultaneity bias by conducting a standard Hausman test for endogeneity of regressors (see Hausmann 1978 and
1983). They strongly accept the null hypothesis of no simultaneity bias for the DEM/USD, the JPY/USD and the
GBP/USD exchange rates. Although the focus and the context of their analysis are very different from what we
investigate in this paper, their acceptance of the no simultaneity bias hypothesis also implies that the estimations
presented in this paper are free of simultaneity bias.