# Reproduce the Australian Spending Report 2026 (v1.2.0)

Report: https://kangamall.au/research/australia-spending-map/

Unzip the package into a new folder. Keep its `docs/research/007-consumer-spending-map/` structure. Run commands from the folder containing `docs`, not from inside the study directory. No access to Kangamall's private repository is needed.

1. Use Python 3.12.14 with `openpyxl==3.1.5` and `matplotlib==3.11.2` installed in your own virtual environment. The source manifests record the original runtime.
2. Create `.local/research-spending-map/raw/` inside the unpacked folder.
3. Download `5204042_Household_FCE.xlsx` and `TSADC04.xlsx` from the exact URLs in `data/manifest.json` and `data/tourism-manifest.json`, and save them in that raw folder. Do not alter the files. Extraction checks the recorded SHA-256 values and stops if ABS has replaced a file; obtain the matching revision or explicitly review the revised data instead of bypassing the check.
4. Run:

```sh
python docs/research/007-consumer-spending-map/analysis/extract.py
python docs/research/007-consumer-spending-map/analysis/build_report.py
python docs/research/007-consumer-spending-map/analysis/verify.py
python docs/research/007-consumer-spending-map/analysis/review_release.py
```

`extract.py` recreates the CSVs and manifests, `build_report.py` recreates the Markdown report, figures and claim/source registers, `verify.py` checks the original XML cells and repeatable extraction, and `review_release.py` independently recomputes derived tables with Decimal arithmetic. Use `--refresh` with the last command to check current upstream hashes again; this intentionally fails if the source revision changes.

Verify the supplied files against `SHA256SUMS.txt` before regenerating. Paths in that checksum list are relative to `docs/research/007-consumer-spending-map/`. Numerical CSVs and rendered chart values should match; Python/library/platform changes can affect output bytes or image rendering. On systems using LF-only text files, extraction writes CSV records with the standard CSV CRLF line endings before verification.

Do not use Python's `-O` flag: the analysis uses assertions for source and calculation checks. No personal or respondent-level data is included. Read `methodology.md`, source limitations and `REUSE.txt` before using results. Technical cross-checks are not independent human peer review.
