Art and science: JobKeeper and the costing of public policy (part II)

20 July 2020

Michael Di Francesco presents a two part series on the costing of public policy using the case of the COVID-19 JobKeeper program. In this second piece, he surveys some key methods of policy costing in Australian government, with a focus on the critical role of behavioural assumptions and the exercise of professional judgement. Costing is often a flashpoint for policy argument, and emphasising policy logic rather than mechanistic precision in costing may generate significant benefits for policy-making.

Costing is one of those ‘taken for granted’ technical practices, widely perceived as producing mechanistic input to decision-making. It is also a critical building block of multiyear budget frameworks widely used by governments to account for the future consequences of current year spending decisions. Despite their strategic aspirations, in practice such frameworks emphasise iterative budget processes to reconcile spending constraints and program costs. But even when the frameworks are operating in routine policy environments the processes and outputs of costing are in constant tension between science and art.

Costing practices inform central policy processes in Australian government in several ways: operating rules govern the costing of new policy proposals (NPPs) in the cabinet Budget process, and policy costing guidelines (PCGs) issued prior to elections by Treasury and Finance frame the costing of policy commitments. Both practices emphasise the role of implementation assumptions in explaining policy. In most instances, programs providing income support or services will be governed by statutory eligibility rules. These rules, which define policy interventions, are usually designed around target groups, and cost estimates reflect projections of target group demand.

Policy design and demand estimates are therefore largely a function of the credibility of behavioural assumptions. Highly specialist expertise can have a role to play here, for example the project specific input of BETA, the behavioural economics team in the Department of the Prime Minister and Cabinet. However, it is routine for behavioural assumptions to be specified by officials in proposing agencies based on their experience and knowledge of the logic and likelihood of policy interventions. In either case, anticipated direct behavioural responses will often involve the exercise of professional judgement. Good costing will require transparency around the assumptions used and the confidence intervals applied. Reliable information is not always readily available in routine policy environments, so the shadow of uncertainty is even darker in crisis conditions. In the case of JobKeeper, as the impacts of COVID-19 unfolded in real time, the absence of data and precedent played havoc with Treasury efforts to forecast business closures and unemployment, both of which loomed large in venturing demand projections.

Drawing all this together, policy costing is reliant on three elements: the forecasting capacity of central economic agencies; the availability and quality of policy-specific information that resides with the agencies proposing policy; and the reasonableness and consistency of behavioural assumptions. These elements suggest that, contrary to popular perception, costing is in equal measure art and science. Yes, costing operates within broadly accepted methodologies, but the final product is routinely a negotiated outcome that reflects the application of contending professional judgements and mutual adjustment.

What we can take away: Target assumptions, not exactitude

Costing is a necessary but insufficient tool of policy practice. The JobKeeper costing experience is a reminder that flaws reside within contemporary fiscal frameworks. By design these frameworks use cost estimates to discipline budget and policy decision-making. But, in practice, such estimates are habitually regarded as concrete fiscal targets and are, increasingly, ‘weaponised’ by governments and oppositions alike. Rather than broadening the basis for informed policy deliberation, the estimates of costs can be subject to a type of reification that narrows and closes policy and political debate. This prompts two closing observations on the status of costing and the potential for clarifying its role in policy process.

Temper expectations

The record of budgetary reform in Australia confirms how the search for ever more sophisticated cost methodologies can overwhelm human and systems capacity without necessarily delivering better information. Observation one is that we should temper expectations about what costing is, and what it can deliver. Usable cost information is of course essential to budget and policy decision processes, but it should be accompanied by more modest aims to reinforce awareness of the resource implications of decisions rather than deliver unattainable mechanistic exactitude. In this vein, there is much to be said for the adage, attributed to John Maynard Keynes, that it is ‘better to be roughly right than precisely wrong’.

Open up policy assumptions

Once decisions are taken, and policy (or intended policy) announced, there is a strong argument for privileging transparency, especially in the context of crisis response. What might transparency look like? This constitutes observation two: given what we know about the art and science of costing, it could conceivably involve mandating full release of the evidentiary and analytical bases of declared policy for the purposes of policy deliberation, improvement, and accountability. What specific behavioural assumptions were made? What empirical research and data were available, and how was it used? What logics informed policy design thinking? In short, it might emulate the analytical approach adopted by the Parliamentary Budget Office as an open costing transparency process. Indeed, such a process might assign a more prominent and expanded role for the Office as an independent reviewer of costing for declared policy, one less tethered to the electoral cycle, and operative in times of both routine and non-routine policy environments. Policy costing can be a flashpoint for policy conflict. Arguments about cost are inevitable, but substituting assumptions for exactitude as the focus of such debate may be one way of extracting much more value from the policy process.

Filed under:

Updated:  28 September 2020/Responsible Officer:  Crawford Engagement/Page Contact:  CAP Web Team