Statement on Monetary Policy – November 2024Box D: Annual Review of the Forecasts

Each year, a review of the RBA staff economic forecasts is undertaken to assess what we have learned about the economy and the forecasting approach, with a view to continuous improvement.

Compared with forecasts in the November 2023 Statement, underlying inflation and the unemployment rate in Australia have evolved broadly in line with expectations.[1] However, inflation outcomes more broadly over the past two years have been higher than expected, and quarterly outcomes for underlying inflation since late last year point to relatively slow progress in disinflation. Against this backdrop, we revised down our assessment of full employment, which is the level of employment that can be sustained without generating inflationary pressures, relative to a year ago.

GDP growth has proven significantly weaker than forecast a year ago, driven by weaker-than-anticipated growth in private demand. Taken together, the weaker GDP outcomes, persistence in underlying inflation and the rate of unemployment coming in broadly as forecast suggests that the economy’s supply capacity has been less than projected a year ago. Specifically, measured labour productivity has proven weaker than expected.

Underlying inflation and the unemployment rate have evolved as expected.

Underlying inflation was broadly as expected by staff a year ago, though a bit higher than expected by market economists at the time (Graph D.1). The November forecasts were for gradual disinflation over the following year; however, the December 2023 outcome for underlying inflation was weaker than expected and there has been slow progress in disinflation since then.

Graph D.1
A line graph showing actual year-ended trimmed mean inflation, another line showing its forecast from the November 2023 SMP, and a shading of the range of forecasts between the February 2023 SMP and the August 2024 SMP. The chart shows that actual trimmed mean inflation initially declined a little faster than expected at the November 2023 SMP but since has been quite close to what was expected. Actual trimmed mean inflation and the November 2023 forecast were at the top of the range of forecasts since the February 2023 SMP.

Further, underlying inflation over the past year has tended to come in a bit higher than suggested by our suite of inflation models, which use variables such as estimates of the unemployment or output gap to explain inflation. Staff had imposed upward judgement to the forecasts using the near-term projections for some of the inflation components. This judgement also allowed for a bit more persistence in inflation than the models suggested. Inflation outcomes being a bit stronger than can be explained by the aggregate models contributed to the reassessment in recent months that the level of full employment was lower than previously thought (see the section on full employment below).

By contrast, headline inflation declined more rapidly than forecast in the November 2023 Statement. This was largely due to the expansion of the electricity rebates by federal and state governments announced around May 2024; the forecasts had assumed that the electricity rebates introduced in 2023 would be unwound in 2024, as was legislated at the time. An increase in rent assistance was also introduced. Together, these policies accounted for a large share of the forecast miss over the year. Weaker-than-expected fuel prices also subtracted a little more from headline inflation over the year.

The unemployment rate increased in line with expectations over the year to the September quarter (Graph D.2). The labour market also adjusted to subdued growth in economic activity via a further decline in vacancies and average hours worked. However, growth in employment and labour force participation were both stronger than expected (see the section on employment below).

Graph D.2
A two-panel line graph showing the November 2023 forecasts and actual outcomes for the unemployment rate and average hours worked. The left-panel shows that the unemployment rate has gradually risen and is very close to what was expected at the November 2023 SMP. The right-panel shows that average hours worked initially declined more steeply than expected at the November 2023 SMP and remains lower than was expected at that time.

Domestic GDP growth was considerably weaker than forecast.

GDP grew by 1.0 per cent over the year, compared with expectations of 1.8 per cent at the November Statement. The outcome was also weaker than Treasury forecasts and more in line with market economists’ expectations. In per capita terms, the forecast miss was even larger, given population growth was stronger than had been assumed. The momentum in GDP growth has also been materially different to expectations; GDP growth is yet to show signs of a recovery and has remained around 0.2 per cent in quarterly terms since December 2023. At the expenditure component level of GDP, weaker growth in household consumption and exports, which was only partially offset by stronger-than-anticipated growth in public spending, contributed to the forecast miss (Graph D.3).

Graph D.3
A waterfall graph showing the contributions of each expenditure component to the GDP forecast miss over the year to June 2024, based on the November 2023 forecasts. It shows that household consumption and exports contributed to the weaker-than-expected outcome for GDP, partly offset by stronger-than-expected growth in public demand.

Growth in household consumption was notably weaker than expected over the year to June 2024, with much of the forecast miss related to the weak June quarter outcome (Graph D.4). The level of consumption by Australian households since the pandemic was revised up significantly in the March quarter GDP release and in the annual National Accounts, with Australians spending more on domestic and overseas travel and imported digital services than previously thought.[2] Since late 2023, the response of consumption growth to the stabilisation of real incomes has been weaker than expected and the expected timing of the pick-up in consumption has been pushed out.

Graph D.4
A two-panel line graph showing the November 2023 forecasts and actual outcomes for household consumption and saving. The left panel shows year-ended growth in household consumption which was forecast to gradually pick up over 2023/24 but has remained subdued. The right panel shows the household gross saving ratio that currently sits a little above its forecast level.

Exports growth has also been much weaker than expected over the past year, primarily driven by services exports. Tourism exports remain around 25 per cent below 2019 levels, which was weaker than staff forecasts and expectations from many domestic tourism operations in liaison. Education exports were also notably weaker than expected. This reflects an unanticipated tightening in student visa requirements by the Australian Government over the year and an unexpected decline in the average expenditure of international students.

By contrast, growth in government spending has been stronger than expected, driven by public consumption. The staff forecast of public consumption has been revised higher several times over the past year in response to announcements of additional spending (the staff forecast does not typically incorporate expenditure that has not been budgeted). A large share of the announced additional spending has been for existing services as well as cost-of-living support packages.

Despite weaker GDP, employment growth has surprised to the upside.

Although the level of GDP is lower than previously anticipated, employment growth has surprised to the upside. Labour supply has also increased by notably more than forecast, such that unemployment rate outcomes have been largely as expected.

The unanticipated strength in labour supply reflected stronger-than-expected increases in both the working-age population and labour force participation (Graph D.5). Population growth was materially higher than assumed a year ago, mostly due to fewer-than-expected departures (i.e. many temporary migrants stayed in Australia longer than expected). The participation rate also continued to increase, against expectations last November that it would ease a little over the year. In part, this may have reflected the ongoing structural increase in labour force participation for women and older workers. However, given that the participation rate continued to increase across most demographic groups, it may also have been facilitated by continued labour demand and driven by cost-of-living pressures.

Graph D.5
A two-panel line graph showing the November 2022 and November 2023 forecasts and actual outcomes for the participation rate and year-ended growth in the working age population. The left-panel shows the participation rate has increased by significantly more than was expected in either 2022 or 2023. The right-panel shows that working-age population growth has been significantly stronger than expected in either 2022 or 2023.

Lower-than-expected labour productivity implies the economy’s supply capacity is lower than previously thought.

The rebound in labour productivity growth over the past year was much less pronounced than assumed in November 2023 (Graph D.6). This is consistent with underlying inflation having held up largely as expected, despite weak growth in economic activity.

Graph D.6
A line graph showing an index of productivity growth since 2015 against index forecasts from the November SMP of each year between 2015 and 2023 inclusive. The index shows that the level of productivity remains around the same level as it was in 2016, and that actual productivity growth has been significantly weaker than expected in 2021, 2022, and 2023.

Explaining trends in productivity outcomes has been an area of focus. Part of the explanation comes from a reallocation of employment growth to lower productivity industries. Around three-quarters of employment growth over the past year was in the non-market parts of the economy such as health and social care, where labour demand tends to be less cyclical and often more aligned with government spending. As measured productivity growth in these areas is lower than other parts of the economy, this was a small drag on overall productivity growth. However, productivity growth was weaker than its historical average across a wide range of market and non-market industries, suggesting other factors have contributed (see Chapter 2: Economic Conditions).

The persistent forecast misses on productivity have generated internal discussions about future trend labour productivity growth, which feeds into estimates of potential output over the forecast period. The current assumption is that annual productivity growth will return to around one per cent in the medium term, noting that strong investment, particularly associated with the rapid adoption of technology across many industries, could lead to higher productivity outcomes. However, this remains a key judgement and the staff will continue to assess its validity. Similarly, many peer economies are experiencing low productivity growth, except the United States, and central banks in those economies have recently lowered their trend productivity growth estimates.

Estimates of full employment are also lower.

The staff’s assessment of how far the economy was from potential output or sustainable full employment gradually evolved over the past year. While the unemployment rate was easing largely as expected, there was a gradual accumulation of evidence that suggested that the level of full employment may be lower than previously thought, such that labour market conditions were tighter than assumed.

This re-assessment, which was outlined in the August Statement, was based on a number of factors:

  • Inflation outcomes had been a bit higher (i.e. there was less disinflation) than could be explained by previous estimates of the degree of excess capacity in labour markets.
  • Wages growth remained high relative to productivity outcomes.
  • Information from business and household surveys and liaison points to significant capacity pressures in the labour market, despite the unemployment rate increasing a little.

While the level of full employment cannot be observed directly or summarised by a single statistic, the current (central) estimate of the unemployment rate consistent with full employment (used in the staff’s forecasting models) is around 4½ per cent.

Staff have uplifted the modelling of spare capacity and provided greater transparency around key judgements and further use of scenarios.

The RBA forecasts draw on a range of models and data, augmented by information from the liaison program and staff judgement. Differences between the forecasts and outcomes (forecast errors or misses) may reflect events that were difficult to anticipate, dynamics in the economy that were not captured adequately by the models, unrecognised structural change, revisions to data or incorrect judgements. Differences can also be quite sensitive to the period over which the forecasts are assessed.

In this respect, regularly reviewing forecasts can contribute to a better understanding of current economic conditions and how the economy is evolving over time. Some of the more persistent or sizeable forecast errors detailed in this review have prompted changes to the staff’s models and a reassessment of some of the judgements and assumptions in the latest set of forecasts. Other forecast errors have highlighted a need for further analytical work and revisions to forecasting models; this work is underway and will inform staff forecasts and analysis in the future.

Specific developments since November 2023 include:

  • The staff have uplifted modelling and analysis of spare capacity in the economy, as recommended by the RBA Review. Compared with a year ago, there is a greater focus on how spare capacity affects the forecasts, the models and key judgements.
  • MARTIN – the staff’s main macroeconometric model – has been further incorporated into the forecasting process. Over the past year, infrastructure has been developed that allows MARTIN’s forecast updates (given new assumptions and data) to be compared with the ‘bottom-up’ GDP growth and inflation forecasts. This provides an additional cross-check on the forecasts, alongside the usual comparisons to forecasts from market economists and other institutions.
  • The staff have also made several improvements to their modelling of variables over the past year, building from considerable evaluation of forecasting models for key variables. These include improvements in the modelling of education exports and key components of inflation, and greater use of the near-term leading indicators of the labour market.

Further, the current forecast review has provided impetus for analysis on a number of dynamics in the economy:

  • The impact of population growth on both demand and supply in the economy. There is work planned on how population growth changes are accounted for in the staff’s current modelling framework for economic activity.[3] This also includes improving the understanding of the effect of population flows on potential output estimates and the impact of large changes in population growth on housing and labour markets.
  • An assessment of the outlook for trend productivity growth based on a better understanding of the historical drivers of productivity outcomes.

Looking ahead, a larger and more comprehensive forecast evaluation project is underway to assess the accuracy of staff forecasts over a longer time period and relative to other forecasters. This broader review – to be completed in 2025 – will also include a more fulsome review of the various key models, with the project to be supported by an upgrade to the staff’s forecast infrastructure.

Endnotes

The analysis is based on the latest data: GDP, productivity and wages data for the year to June 2024, and the labour market and CPI data for the year to September 2024. [1]

This resulted in an upward revision to both consumption and imports, and therefore had no effect on overall GDP. [2]

For example, the MARTIN model does not revise variables such as consumption, employment or income for changes to population. There is merit in considering alternative forecasting of some variables in per capita terms. [3]