What committees asked for

How to read a result

Where the words come from. A recommendation shown against a report on one of the registers is the committee’s own text, taken from the report it tabled — nobody has answered it, so no government document quotes it. So is a recommendation whose response never quotes it: where the government answered a report as a whole, or closed it with the form letter without listing what it recommended, the report itself was read and each recommendation is shown with that fact (571 results). Every other recommendation here is the committee’s text as the government reproduced it in its own response. That is the government’s transcription, not the committee’s original, and it is quoted rather than paraphrased so the difference can be checked against the document linked on every result.

What the government said. 3,161 results carry the government’s own sentence about that recommendation. This site does not translate those into a status — “accepted”, “rejected”, “ignored” — because the wording is the evidence and a one-word label would be our summary rather than their answer. Where the sentence is the passage-of-time template, that is what a form-letter closure looks like, and Methods explains how those are identified.

Whose recommendation it is. Dissenting reports, minority reports and additional comments carry recommendations that are not the committee’s. Treating one of those as a committee recommendation would be a serious misstatement, and it is the mistake this page is most likely to make. Where the text or the section it sits in names another author, the result says so and you can filter on it — 388 of the 5,164 do.

That filter is not a guarantee in the other direction. It finds what names itself: a section headed “Dissenting Report”, or a recommendation whose own words begin “Labor Senators recommend”. A dissenting recommendation that does neither is not caught, so the committee’s means “nothing here identifies another author”, not “verified as the committee’s”. Where it matters, the document is linked on every result.

Where a number is contested, the committee holds it. A dissenting report restarts its numbering, so one report can contain two “Recommendation 1” — the committee’s and a party’s. This site publishes the committee’s under that number, and a party’s recommendation appears only where it carries a number of its own. That means a dissenting recommendation can be missing from this index; it does not mean one can be published as the committee’s. Given the choice, the site would rather omit than misattribute.

What a missing response does not tell you. This site can show that a committee made a recommendation and that no response was tabled. It cannot show whether anything was done about it. A government can act on a recommendation without ever answering the committee — through a budget measure, a regulation, an agency's own decision — and it can answer a committee at length and do nothing at all. The two are independent, and this site measures only one of them.

That gap is not one we intend to close by inference. Establishing whether a recommendation was implemented means reading budget papers, annual reports, estimates transcripts and an agency's own record, one recommendation at a time, and forming a judgement about them. That is journalism, and this index exists to make it possible rather than to stand in for it.

What a missing response does establish is narrower, and still real. The Senate has required a response within three months since 1973 and the House within six since 2010. A report with no response is a failure to account to the chamber that asked for one, whatever else may have happened. That obligation is owed to the Parliament, and it is not discharged by acting quietly. Treat a result here as the beginning of a question rather than the answer to one. If you know a recommendation was acted on, tell us and we will record it.

What is not here. This is not every recommendation an Australian parliamentary committee has made. It is every one this site can show from the documents it holds: government responses tabled since mid-2022, which is as far back as the Tabled Documents system goes, and the reports currently on the two registers. A response that answers in narrative prose without numbered recommendations yields nothing of its own; 227 of the responses in the corpus are like that. For 76 of them the report was read instead, which leaves 151 that this index can show nothing from; a search names them when they match.

The oldest reports are here because someone went and got them. Both halves of this index come from the Parliament’s Tabled Documents register: the government responses, and the PDFs of the reports still waiting. That register holds nothing before 2022, so every report on the two registers older than that — the longest waits on this site, the ones a reader is most likely to look for — was missing, and nothing on the page said so. 25 of them have since been fetched from their committees’ own pages, read by the same parser as every other report and checked the same way; each of those rows records that its document was collected that way rather than taken from the register. One report has not been found yet, and its recommendations are still missing from this index.

There is a second shape it cannot read, and it matters more than the first because the clearest example of a form-letter closure this site has is written in it. Where a response sets out its answers in a table and the numbering is the word processor’s rather than typed, there is no “Recommendation 4” in the text for anything to find, and the document yields nothing at all. The response to the 2002 certain maritime incident report — the children-overboard inquiry, answered in 2026 — is one such document, and searching for it here returns nothing. It is in the Senate’s response record and in the dataset; only its recommendations are missing from this index.

Every recommendation here was located in the extracted text of its own document, under the number it claims, and any row whose words could not be found there was removed. One row did not pass that check and is listed in recommendations_dropped.csv. That is a consistency check, not an independent one: it catches a row assembled from text that is not in the document, or filed under a number printed only on a contents page. It cannot tell whether the split between the committee’s words and the government’s was made in the right place, or whether the author named is the right one. Those are read by hand, a sample at a time, and what that reading finds is published in the corrections log. Covers 23 October 2002 to 27 August 2026. The whole index is in the public dataset, licensed CC BY 4.0. If a recommendation is wrong, write to corrections@toohardbasket.org.au.