5. Possible Explanations for the
Correlation between Net Positions and Exchange Rates
One interpretation of our results is that they reflect a tendency
for speculators to react to currency movements after the fact.
For example, it is possible that speculators tend to go long
(short) after the currency appreciates (depreciates), so that at a
weekly interval, one would find the strong contemporaneous
relationship documented in the previous section. The
economic logic behind such simple trend-following behavior,
though, is hard to accept even with the Granger causality
results for the Canadian dollar and Swiss franc. Specifically, in
efficient markets, exchange rate movements over the previous
minute, hour, day, or week have no new information about
future exchange rate movements. Thus, basic trend-following
behavior does not make economic sense for speculators as a
whole, though more sophisticated trend-following models
indeed drive some speculators. The nature of exchange rate
dynamics would therefore argue against simple trendfollowing
behavior as the source of the measured
contemporaneous relationship between positions and
exchange rates.19
Another explanation for the strong correlation is that the
two variables tend to respond jointly to market developments,
such as the release of public information. For this argument to
hold, exchange rates and net positions need to react at similar
speeds since speculators stop changing their positions once
currency values reflect this new information.20 But given the
nature of exchange rate determination—with rapid price
adjustments possible without the need for trading—it seems
likely that spot exchange rates would react to new information
well before speculators had a chance to change their futures
market positions. In that case, the change in currency values
would mitigate the incentive for speculators to change their net
positions, calling into question the notion that the observed
correlation of net positions and exchange rates occurs because
both move together in response to a third factor.