Show us the models, Mr Abbott

Lobbecke

AS well as being policy fiascos, the carbon tax, the National Broadband Network and the mining tax have this in common: they were based on economic modelling that has been kept hidden from the public.

With serious questions now also arising about the Gonski report's modelling, Tony Abbott should commit a coalition government to a Charter of Modelling Honesty that would bring greatly needed transparency to future policy evaluation.

The collapse in European carbon prices, from about $50 per tonne in 2008 to barely $3 today, highlights the issues involved. Ever since the government linked its Emissions Trading Scheme to the European carbon market, Climate Change Minister Greg Combet has defended the budget's estimate of a $29 per tonne carbon price in 2015-16 as a "Treasury forecast on which the government is entitled to rely".

However, Combet has not provided any information supporting that "forecast", instead pointing to Treasury's carbon tax modelling. But contrary to Combet's assertions, that modelling does not "forecast" a $29 price in 2015-16.

Rather, it calculates that $29 price as the lowest global carbon price in 2015-16 that would be consistent with the government's aspirations for cutting emissions. The $29 figure was therefore not a prediction of likely prices in 2015-16; it was merely the minimum price needed for deep global emissions reductions to occur.

Unfortunately, Treasury has never set the record straight, instead allowing Combet to repeatedly mislead the public. And as the model and data Treasury used have not been released, it has been impossible to test the impacts of much lower world prices on Treasury's conclusions.

The NBN is, if anything, even worse. The government's $30 million implementation study and subsequent reports gathered all the data independent experts would need to conduct a cost-benefit appraisal; but it has never been released. As for NBN Co's own modelling, only the results, rather than the models and underlying data, have been disclosed -- and those results are implausible, with NBN Co assuming, for example, that its costs will decline unrealistically quickly.

Adding to the problems, NBN Co plays fast and loose in its costing methodology. Its report issued on Friday is a case in point. It provides no explanation of how its estimates of fibre rollout costs per premise are derived. As best one can tell, those estimates ignore substantial overheads and do not properly include the network's accumulated losses. But without the underlying model and data, it is impossible to fully identify and correct the errors.

The mining tax adds to this litany of modelling misdemeanours. According to Treasury, its estimates of revenue from the tax were based on detailed modelling (which, again, the government has refused to publish). Perhaps, but there must be substantial doubts as to that modelling's quality, as mining experts always dismissed the government's promised El Dorado as fanciful.

To top it all off, important issues are now emerging about the Gonski report. That report's sins were largely of omission, as it ignored inconvenient facts. It claimed, for example, that "adopting a national curriculum" characterised the world's "high-performing school systems"; but it nowhere noted the lack of a national curriculum in countries such as Canada, whose education system it praised.

The difficulties, however, go much further. Rather, the report was less than frank about modelling it commissioned on per student outlays. That modelling was not released when the report was published; but it was loosely summarised in the report as examining recurrent expenditures at schools where at least 80 per cent of students achieved above the national minimum standard in both reading and numeracy.

What the report did not say, however, was what its own modelling found: that even correcting for student and school characteristics, those schools do not have materially higher outlays per student than schools that perform less well. Poor performance, in other words, is not due to insufficient funding. But that finding, which cuts at the heart of the government's rhetoric and that of the teachers' union, is simply never mentioned in the report.

Regrettably, these are not isolated instances, as a special issue on policy modelling of the Australian National University journal Agenda shows. Little wonder: for just as hypocrisy is the homage virtue is paid by vice, so the Gillard government is every bit as desperate to claim its policies are "evidence-based" as it is to prevent that evidence from being properly tested. And with the government using modelling as agit-prop, it has debased the tool's credibility: bad modelling drives out good, reducing the effectiveness with which policies are scrutinised.

But it would be misguided to think this government's demise would suffice to stop the rot. Canberra is not merely attached to junk modelling: it has become addicted to it, just as it is now addicted to overblown promises, fatuous targets and windy self-justification. After a spasm of propriety, a new government could slip back into the old ways. And it would do so all the more effortlessly as that would be welcomed by the officials, advisers and swarms of consultants who have luxuriated in the quiet life Labor's lack of transparency allows.

An Abbott government should therefore commit to a Charter of Modelling Honesty that sets standards for the comprehensive release of the models and data on which policy assessments are based. That charter could be informed by the Productivity Commission, which has an excellent record in this area.

Good writing, said George Orwell, must be "like a windowpane". So must good policy evaluation. Until it is, it will not protect the public from poor policy, as it can and should, but only defend indefensible policy from the scorn it deserves.