Predicting Finnish economic activity using firm-level data / Paolo Fornaro

The paper presents a new way to compute flash estimates of the Finnish monthly economic activity indicator, using firm-level data. A two-step procedure is used where the extracted common factors from the firm-level data are used subsequently as predictors in nowcasting regressions. It is shown that these factors-based nowcasting models lead to a superior out-of-sample performance compared with the benchmark models including autoregressive and random walk benchmark, even for very early estimates. Moreover the quarterly GDP flash estimates based on factor models provide a timelier alternative to the current estimates, without loss in accuracy. Hence large firm-level datasets are useful in predicting aggregate economic activity in a timely fashion.

Pdf: Predicting Finnish economic activity using firm-level data / Paolo Fornaro