An alternative interpretation of ERM crises is based on second-generation models of
self-fulfilling speculative attacks and multiple equilibria in foreign exchange markets, in
which policy shifts in a more expansionary direction in response to the attack (Flood and
Garber 1984b, Obstfeld 1986). For the ERM subsample, we find little evidence of this
pattern. Thus, while our findings cast doubt on the relevance of first-generation models for
our ERM episodes of speculative crisis, they do not establish that second-generation models
of self-fulfilling attacks necessarily fit the facts.
It is important here to note a problem of observational equivalence.2 While the
absence of differences in monetary and fiscal variables in periods leading up to speculative
attacks and other periods is consistent with models of multiple equilibria, it is also consistent
with a restrictive class of models with unique equilibria. Models like those of Flood and
Garber (1984b) and Obstfeld (1986) generate multiple equilibria and self-fulfilling crises
because they assume a contingent policy process in which policy shifts only in the event of
an attack. One can also imagine a model in which policy is expected to shift in a more
expansionary direction with certainty; the shift is not contingent. Anticipating that
eventuality, speculators may attack the currency just before the policy shift is observed. This
is a model with a unique equilibrium ii which the speculative attack is motivated by
imbalances in underlying fundamentals, but those imbalances only become evident after the
attack. Thus, our results for the ERM, which fall to detect distinctive behavior on the part