In addition, given the uncertainties surrounding our forecasts of the pace of labor market improvement and the degree of remaining slack, monetary policy has to be determined largely by incoming data and the signals that data provide about the health of labor markets. If the economy disappoints we should be in no rush to raise short-term rates, but if the economy improves more quickly than anticipated we should raise short term rates earlier. Thus, we should be moving away from providing date-based forward guidance, and instead focus on what incoming data tell us about reaching full employment and 2 percent inflation within a reasonable time period.
...
In fact, I actually hold the view that as we approach levels of unemployment that many consider “full employment,” the Fed should no longer issue guidance on the approximate timing of any monetary policy changes.
I do not intend this to reduce transparency in monetary policymaking. Rather, I simply want to acknowledge that any reference to calendar dates has the potential to be inaccurate. The date of “liftoff” from near-zero short-term rates is highly dependent on how the economy actually evolves – in other words, is going to be tied to the current and expected path of inflation and employment. We are getting close enough to targets that, given the uncertainty around forecasts of these variables, incoming data that cause Federal Reserve policymakers to significantly change our outlook for the economy will shift any expected lift-off date forward or backward in time. So, again, reference to calendar dates as we approach targets has the potential to be inaccurate.