The International Monetary Fund published a working paper presenting how a Bayesian Structural Time Series (BSTS) model can be used to nowcast quarterly economic growth in Tanzania using a range of high-frequency economic indicators, aimed at improving short-term growth monitoring amid external volatility and limited timely data. The paper describes a BSTS framework that combines trend, seasonal, and regression components, and uses spike-and-slab variable selection to identify the most relevant indicators. It sets out an approach to model selection and evaluation, including robustness checks and sensitivity analysis, and reports the model’s relative performance. The analysis also highlights the model’s ability to extend to longer forecast horizons and incorporate dynamic regressors to track growth patterns in changing economic conditions.
International Monetary Fund 2026-03-20
International Monetary Fund working paper applies Bayesian structural time series model to nowcast Tanzania’s quarterly growth
The International Monetary Fund released a working paper on using a Bayesian Structural Time Series model to nowcast quarterly economic growth in Tanzania with high-frequency indicators. The paper details the model's framework, including trend, seasonal, and regression components, and highlights its ability to extend forecast horizons and incorporate dynamic regressors. The approach includes model selection, evaluation, and robustness checks.