Nowcasting methodology

Image
NERO Methodology paper front cover

What is nowcasting?

Nowcasting is an emerging methodology that typically uses both traditional and real-time data, as well as big data techniques, such as machine learning, to estimate current trends in a timely manner. This alternative approach to labour market data analysis can provide insights that are more timely, are available more frequently, and at more granular levels.

This dashboard visualises the output from developing experimental monthly nowcasting estimates of the occupational (ANZSCO 4-digit) employment by region (SA4) to support policy makers, educators and labour market participants to make more informed decisions.

The National Skills Commission’s nowcasting employment estimates are experimental (Beta version) and subject to revision. Given the experimental nature of the data and the large number of series predicted by the model, some estimates may be volatile and subject to error. Caution should be exercised when using the data, particularly for smaller series and where large movements in the estimates are observed.

The estimates produced by the National Skills Commission are broadly consistent with the publicly available Labour Force Survey data produced by the Australian Bureau of Statistics. This includes the Labour Force Survey estimates for total employment by SA4 region and for employment by ANZSCO 4-digit occupation. Some differences between the ABS Labour Force Survey and NERO estimates may emerge at aggregate levels due to various smoothing and analytical techniques that have been applied as part of the NSC’s modelling process. Any differences are coincidental and should not be interpreted as a differing view of the underlying trends or estimates of employment for those series.

Methodology

For more details, please see the methodology paper at the bottom of the page.

Supervised machine learning techniques such as random forest, gradient boosting and linear regressions were used to train models to find patterns that can be used to predict the “now”. Many variations of models were tested and combined into a stacked model.

A range of data sources were used as modelling inputs. These include:

  • ABS Labour Force Survey Total employment by region
  • ABS Labour Force Survey Occupational employment
  • ABS Census Occupational employment by region
  • ABS Weekly Payroll Jobs by industry and region
  • ABS National Accounts Gross State Product
  • NSC Internet Vacancy Index Online job advertisements by region and occupation
  • Burning Glass Online job advertisements by region and occupation
  • DESE jobactive job placements by occupation and region
  • Home Affairs Skilled Visa holders by occupation and state/territory

Note that a minimum value of 10 has been applied to all estimates.

The following ANZSCO 4-digit defence related occupations have been excluded from these nowcasting estimates:

  • 1391 - Commissioned Officers (Management)
  • 1392 - Senior Non-commissioned Defence Force Members
  • 4411 - Defence Force Members - Other ranks

Download the methodology paper

Creation Date