This article explores the use of diffusion indexes to forecast macroeconomic time series variables using a large number of predictors. The authors propose an approximate dynamic factor model as the statistical framework for estimating the indexes and constructing forecasts. The method is applied to forecast eight monthly U.S. macroeconomic variables using 215 predictors over a 1970-1998 period. The forecasts are compared with univariate autoregressions, small vector autoregressions, and leading indicator models. The results show that the diffusion index forecasts, based on a small number of factors (usually one or two), outperform the benchmark forecasts, with improvements becoming more significant at longer horizons. The study also highlights the importance of using a small set of driving variables to capture the dynamics of macroeconomic time series.This article explores the use of diffusion indexes to forecast macroeconomic time series variables using a large number of predictors. The authors propose an approximate dynamic factor model as the statistical framework for estimating the indexes and constructing forecasts. The method is applied to forecast eight monthly U.S. macroeconomic variables using 215 predictors over a 1970-1998 period. The forecasts are compared with univariate autoregressions, small vector autoregressions, and leading indicator models. The results show that the diffusion index forecasts, based on a small number of factors (usually one or two), outperform the benchmark forecasts, with improvements becoming more significant at longer horizons. The study also highlights the importance of using a small set of driving variables to capture the dynamics of macroeconomic time series.