Forecasting Inflation in Morocco with Parsiminous Factor Augmented Shrinkage Methods using Big Data

In this paper we aim to determine and select crucial factors and variables to predict future inflation rates of Morocco using principal component analysis and the elastic net method of Zou and Hastie (2005). We make use of a large data set on inflation rates of several African countries. We therefore construct several "hybrid” forecasting models using static and dynamic regressor parameters to conduct an out-of-sample forecasting experiment. The results of our empirical study reveal that all the factor-based forecasting methods, whether it has static or dynamic parameters, outperform non-factor-based methods including a benchmark autoregressive model for different windowing methods. The inclusion of time-varying parameter drastically improves the forecasting performance of all our forecast models.