The closer we get, the better we are?

Not always. A general requirement in models of expectations formation is that forecast accuracy should improve when getting closer to the target date. We apply a multi-horizon bootstrap test to a unique Israeli survey of firm-level inflation forecasts, which encompasses periods of three digits annual rates, and periods of moderate and low inflation. We find that improvements in forecast accuracy were significant during high and moderate inflation periods, but not in the low-inflation environment. The results are in line with rational inattention, induced by a decline in inflation predictability. Further support is provided by evidence from the US SPF survey.