A note on the representative adaptive learning algorithm

We compare forecasts from different adaptive learning algorithms and calibrations applied to US real-time data on inflation and growth. We find that the Least Squares with constant gains adjusted to match (past) survey forecasts provides the best overall performance both in terms of forecasting accuracy and in matching (future) survey forecasts. •Learning algorithms are assumed to represent agents learning-to-forecast behavior.•Main algorithms in the literature: Least Squares (LS) and Stochastic Gradient (SG).•We compare the forecasts associated with the LS and the SG algorithms.•We use US real-time data on inflation and output growth.•Our results favor the use of the Least Squares algorithm as representative.