Tuesday, 29 November 2016

Analysis of Australian government spending across all states (part 2)

After much scrounging around, I have collected a full dataset of government spending across all states/territories and the federal budget.

The results are interesting. I was initially motivated to collect this data because of a debate on Facebook where I argued that Australia spent more on its military than on education. It turns out that's true at a Federal level but not when state spending is factored in (you were right Erik and Claire ;)).

The next step for me is to turn this data into an interactive visualisation using D3.js.

Some notes on the process of collating data:
There are some interesting tidbits in the budget papers
The QLD budget features some strange line items.
"Department of Aboriginal and Torres Strait Islander Partnerships Investing in Success" - $0 every year shown
"Domestic and Family Violence Response - Specialist Domestic and Family Violence Courts" - $0 every year shown
"Maintenance Funding for Critical Projects" - $0 for every year shown

Budget measures data is quite difficult to find in a usable form
My lack of financial au fait made it a bit hard to trawl through the budget reports to find what I need. The dense nature of the budget papers is probably helpful for some purposes but I craved spreadsheets. There were some CSVs on data.gov.au but they were not useful most of the time. The only spreadsheet that I found usable was from Victoria. I ended up having to go to the Bill of Appropriation for most of the states as that provided summary data on the total budget as well as the appropriations for that year.

Categorising spending was a bit tricky
The budget datasets came at different levels of abstraction. In some cases (e.g. NSW), department level data was easily available but in others, only the GPC (governmental purpose category) data was available. This made it tricky as I wanted to be able to report on spending on public transport vs roads but the lack of granularity made this impossible.

I was also a bit unsure about how to categorise some line items. For example, should the State Library of NSW fall under Education or "Recreation and Culture"? (I ended up putting it under "Recreation and Culture" as it seemed to belong with cultural institutions like the Museum of Applied Science).

Infrastructure spending was particularly tricky. In some datasets, infrastructure had its own line items but in other cases it was reported under the relevant department (e.g. transport/energy).

Had to avoid double counting
In the NT and ACT, there were quite a few line items where federal contributions were listed. To avoid double counting, I did not include this spending as it was included in the federal analysis.

List of data sources

ACT http://www.legislation.act.gov.au/b/db_51822//
NT https://budget.nt.gov.au/__data/assets/pdf_file/0011/277616/BP1-Speech.pdf
WA http://static.ourstatebudget.wa.gov.au/16-17/factsheets/key-agencies.pdf?
SA https://www.legislation.sa.gov.au/LZ/B/CURRENT/APPROPRIATION%20BILL%202016/B_AS%20INTRODUCED%20IN%20HA/APPROPRIATION%20BILL%202016.UN.PDF
NSW http://data.nsw.gov.au/data/dataset/nsw-budget-paper-2/resource/0b1ab94e-10a1-4fa9-b483-1cd72f5df778
VIC https://www.data.vic.gov.au/data/dataset/state-budget-2016-17-consolidated-government-purpose-classification-data-general-government-sec/resource/8159bd99-48f9-4a49-afb0-b8db55a64925
QLD https://www.legislation.qld.gov.au/Bills/55PDF/2016/B16_0067_Appropriation_Bill_2016.pdf

No comments: