The article presents a method to predict GDP growth in real-time using various economic and financial indicators. The researchers considered different models with constant or changing variances and regression coefficients. By using Bayesian methods, they were able to estimate the model parameters and generate predictive densities. The results show that the proposed method accurately forecasts real-time GDP growth in the U.S. from 1985 to 2011, comparable to other econometric methods and survey forecasts. The model with stochastic volatility was particularly useful for reliable density forecasts.