This paper assesses to what extent forecasters make efficient use of competitors' forecasts. Using a panel of forecasters, I find that forecasters underuse information from their competitors in their forecasts for current and next year's annual GDP growth and in ation. The results also show that forecasters increase the attention to their competitors as the forecast horizon decreases. In a model of noisy information with fixed target forecasts, I confirm the empirical results of underuse of competitors' information. I also extend the model to include a revision cost and show how this can explain the observed inefficiency and observed horizon dynamics. Using the same model framework, I also rule out overconfidence as the main explanation of the observed behavior.