摘要：When policy forecasts are based on the policymaker's present and past actions,
current policy affects expectations of future policy, contrary to what happens
when forecasters can replicate policymaking perfectly. We show that when
forecasts are generated through any linear combination of present and past
policy functions that produces expectations consistent with the implemented
policy, the optimal discretionary policy exploiting learning converges toward
the optimal commitment plan as we approach a situation where people do not
discount the future. Since influencing expectations permits improving policy,
successful policymakers need to know how policy expectations are formed and how
they can affect these expectations.
