Convergence Rates of Agents' Learning in Macroeconomic Models