Published by the Australian Financial Review, Monday 29 May

Treasury’s modelling of the economy is rarely front-page news. But the gory details matter. A lot.

The Commonwealth Treasury has typically over-estimated future budget balances by around $30 billion a year for the past seven years. And this estimate has excused successive Treasurers from making the really tough decisions required for budget repair.

Why has Treasury been wrong in the same direction seven times in a row? And why has it been so slow to adapt when reality continuously failed to fit the theory?

Although prediction errors are inevitable, Treasury’s goal should be “unbiased” predictions: happy accidents should be as likely as downside surprises. So it is worth understanding why Treasury has lucked out seven years in a row.

First, Treasury systematically overestimated nominal GDP growth. In part it misjudged the underlying economy – in particular it assumed a big pick-up in non-mining investment. And in part Treasury did not foresee that commodity prices would fall so sharply from their 2012 peak.

Second, wages growth has been surprisingly slow, particularly given modest unemployment. The breakdown in this traditional relationship has caught Treasury – as well as other domestic and international forecasters – flat footed.

Tax collections

Third, the models that translate economic predictions into tax revenues also seem to suffer from optimism bias. Treasury has predicted a strong recovery in company tax and capital gains tax receipts since 2011. Either losses sustained in the GFC are much larger than anyone realised, or the Treasury models no longer reflect the reality of how profits and asset price rises translate into tax collections.

The biggest underlying issue is that Treasury uses a somewhat formalistic model for projecting the economy three to four years out from budget night. It assumes the economy will return to trend, over a five-year cycle. And if the economy is currently below trend, it assumes that the spare capacity will be used up within the cycle – so that there will be a period of above-trend growth within the four-year estimates period.

This invites a lot of questions. Treasury is still using the history of the past 30 years to assume that “trend” growth is much higher than the actual growth for developed economies since about 2005. Treasury is still assuming that the economic cycle is only five years long, even though the current cycle is clearly much longer than that. And Treasury is still assuming that spare capacity will get used up, even though an “output gap” (as it is known by specialists) has persisted in many European countries for at least a decade.

In practice, the projection model is a sophisticated factory for generating unjustified optimism when the economy is growing slowly.

This is exemplified by the wages forecasts. While actual wages growth has sunk lower over the past four years, projections for future wage growth have risen higher. Treasury is now projecting wages will grow in 2020-2021 at 3.75 per cent – about the same level as at the height of the mining boom.

Plausible downside

If Treasury’s projections were unbiased, then a plausible downside to this projection would be as likely as the equivalent upside. If wages stay at their current levels – not a great outcome, but all too plausible – wages would only grow at 2 per cent in 2020-21. The corresponding upside to the projection would be wages growing at 5.5 per cent in 2020-21 – and there aren’t many betting on that outcome.

This matters a lot to budgets. Wages growth translates into income taxes. If wages instead continued to grow at 2 per cent, income tax collections would be $12.2 billion lower in 2021, wiping out the projected surplus.

Budget outcomes translate into budget strategy. Successive Treasurers said they were on track for surplus. Which Treasurer would make politically tough decisions that are only necessary if the projections turn out to be wrong?

Treasury has changed things at the margin. It relies more on market input and judgment in estimating future commodity prices. It has pulled back the expected trend growth rate by 0.25 per cent. But fundamentally, too little has changed.

Of course, Treasury has reviewed things. Reviews in 2012, 2015 and 2017 all recommended that Treasury develop an economy-wide model to feed into its forecasts. Such a model is not yet operating. The model proposed in the review released by Treasury last week would replace the mechanical (and systematically optimistic) assumptions of the projection model with forecasts based on historical evidence.

Forecasting agencies

Replacing projections with forecasts would bring Treasury in line with the approach of forecasting agencies internationally. It would also take away the fiction that Treasury is not accountable for the numbers in years three and four because they are “just projections”.

But whatever the model, better outcomes will require more judgment. A model is only as good as its inputs. If the numbers coming out of the model look “off” because the world has changed, then Treasury officials should be willing and able to override them.

This debate has become too important for us to simply trust those operating the machine. A much more substantial review, demonstrably independent of Treasury, is needed to determine how Australia’s budget projection methodology should change.

Otherwise, future Treasurers will continue to hide behind optimistic projections to justify inaction. And meanwhile, the debt burden for future generations will keep growing.