Quarterly Estimation of GSP

This repo delves into estimating and nowcasting Australia’s Gross State Product (GSP), with a focus on transparent data sourcing, and mixed-frequency indicators.

Economic activity at the state level arrives at different frequencies and with revisions, yet decisions often can’t wait for the “final” quarterly releases. This project provides a minimal scaffold to:

• align mixed-frequency indicators,
• run a deterministic, restartable pipeline,
• export figures and tables for reporting,
• and keep the whole process easy to implement or extend.

Project structure

• R/ — R helpers and pipeline code (targets).
• _targets_DataSourcing.R — targets pipeline config and plan.
• data/ — input data files.
• data/schema_mfbvar.csv - describe series metadata and help keep indicator handling consistent
• images/ — generated figures.
• Estimation_GSP.qmd, - Project presentation
• GSP_report.qmd - Project report
• Data_sourcing_GSP.qmd — A quarto document where the data pipeline is executed.
• QGSP.ipynb — The notebook contains data preprocessing, Model fit and EDA scripts.

Data Pipeline

library(targets)
tar_make()

This executes the workflow defined in _targets_DataSourcing.R. Intermediate artefacts are stored under data/, and charts are written to images/. You can visualise the pipeline with:

tar_visnetwork()

An extensive instruction is provided in Data_sourcing_GSP.qmd. Follow the notebook for further context.

The pipeline is declarative, rerunning tar_make() only rebuilds what changed.Data acquisition steps (when included) are scripted so sources and transforms are traceable. If you modify inputs (e.g., add a new indicator to a registry/schema), just update the relevant CSV / function and rerun the pipeline.