Got a policy suggestion? Well, how about we test it and see if it works?

One of the new mantras among policy wonks has been "evidence-based policymaking" - the notion that policy ideas should stand or fall on the basis of research and trials, rather than opinion polling and supposition.

Overseas, this has led to some startling discoveries. Young driver education programs, once thought to reduce road deaths, actually turned out to increase them - by encouraging high school students to drive at a younger age. Moving to Opportunity, a US program that provided housing vouchers for poor people to move out of ghettos, dramatically improved the health of children and their parents. And studies on class sizes have cast doubt on earlier assertions that across-the-board reductions boost students' test scores - as a recent report by the Centre for Independent Studies has shown.

The lesson is that policies, like medical interventions, can be put to the test - saving millions of taxpayer dollars, and improving the quality of government. To be effective, evidence-based policymaking relies on policy trials, which simulate the randomised conditions of a laboratory experiment, and access to high-quality data. Unfortunately, both are largely absent in Australia.

As the NSW election campaign has demonstrated, politicians are about as ready to engage with policy trials as with redheaded fishmongers migrating in from the north. The parties, it seems, are big on rhetoric but not on putting their ideas to the test. If the Coalition believes that Parenting Partnerships will reduce conflict in schools, they should propose a one-year experiment - randomly implementing them in 100 schools and reporting on whether schools with partnerships have better test scores and retention rates. If the Greens believe that dispensing heroin is the way to go, they should suggest comparing the outcomes of a group of addicts who are eligible for it with a group who are not. And instead of watching incarceration rates skyrocket, Labor could trial and test training and rehabilitation policies in different jails, to see which does best at helping ex-cons find jobs.

The only example of evidence-based policymaking that we are aware of in NSW was last year's Drug Court evaluation. Carefully administered, the research has provided powerful evidence that this court provides a more cost-effective solution than the traditional judicial system. At least it's a start.

Why don't we see more randomised trials in Australia? One impediment is a cultural attitude that government services are an entitlement, and therefore must not be rationed. Yet it is time that this conventional wisdom was balanced against the benefits that can flow from careful pre-testing of government programs.

Even as Australians have started to embrace testing, our institutions have failed to follow, denying access to data, or imposing hefty fees. By contrast, the United States statistics bureaus apply a simple rule to their data: if the public answered the questions, the public have the right to analyse the data. And these inputs sustain a proliferation of thinktanks that debate policies based on outcomes, rather than conjecture.

But in Australia, the picture is transformed. The Australian Bureau of Statistics makes virtually no data of any complexity freely available. Vast stores of intriguing data are aggregated into bland facts for publication in the yearbook, rather than released for primary analysis. When researchers cannot track individual education, health, crime and labour market experiences, we lose the ability to make subtle judgements about policy effectiveness.

Charging for statistical data is a policy that is hard to rationalise. Simple economics tells us the price that should be charged for "public goods" - such as clean air, street lighting, or national defence - is zero, otherwise these public resources will be underused. In the case of data, there is an extra public benefit: good research leads to better public policy.

As much as it pains two economists to say it, we are talking about a free lunch. Leading researchers from around the world are naturally drawn to the best and most detailed data. The ABS currently collect the best data - the problem is that they will not release it. It is a sure bet that if the ABS starts to release these data, then top researchers - including our colleagues at Harvard and Stanford - will start to analyse them, providing free insights to Australian policymakers.

Hence we offer this twin challenge: First, Australia's federal politicians - of all political persuasions - should commit to providing the ABS with the extra $7 million required to abolish data access fees, and commit to opening up the databanks. Second, new policy proposals should be subject to random trials before being funded. The cost of policy mistakes is surely greater than that of small-scale random trials. And NSW should take the lead.

To those who don't sign on, we say: "chicken!". One can barely disagree on cost grounds. Rather, the fear must be that with real evidence, voters might discover that reality does not match political rhetoric.