Appendix 1: Technical notes on forecasting methodology

The Department of Labour uses a Computable General Equilibrium (CGE) model of the economy (developed by BERL) run in-house to forecast changes in GDP. For each industry group, which takes into account overall production constraints (such as the size of the labour force) a wide range of macro-economic variables (such as international demand for our goods) and the interaction between growth in each industry and the impact of that growth on other industries. These GDP growth forecasts are then combined with estimates of future productivity growth in each industry to forecast industry employment growth. Following this, historical changes in occupational shares within industries are used to project estimated future changes in occupational employment. Change in qualifications shares for each occupational group is then used to make qualification based occupational employment forecasts.

In essence this is a top-down economy-wide approach, where pertinent macro-economic data and informed assumptions are fed into a model that generates employment projections. Often these models are used in conjunction with bottom-up information, for example, through consultation with industry groups and other government agencies. This bottom-up information is then fed back into the model to generate results that are more plausible and consistent with on the ground experience. The availability of this bottom-up information across different industry groups has not been possible in the Department's modelling effort to date, with some exceptions (e.g. primary sector) but may become more important in the future.

The nature of the employment demand modelling is that the projections by industry group are expected to be most reliable, as they incorporate the most direct link to the Industry GDP projections. The occupational projections require a further assertion about the stability of the changes in occupational shares within each industry, and the qualification projections a further extension of this again. While it is difficult to measure the impact of these factors on the likely variability in the forecasts or 'noise' as it is not something that is simple to quantify (such as a sampling error), it might be useful for the reader to bear this in mind.

The labour supply model estimates the likely supply of the labour force by gender, age cohorts and highest qualification levels (used as proxy for skills) that is likely to be available in the future. This model is primarily based upon demographic projections of population growth, informed by mortality, fertility and net migration assumptions. It uses this population projection, together with projected data on age, qualifications and gender to estimate broad labour supply by qualification.

Critical factors in the Department's labour demand and supply modelling

Retirement rates

Retirement rates are derived from job openings arising from the "turnover" in existing jobs, due to people aged 50 years and over leaving the labour force. The baby boom generation has approached their retirement years and as a result, the age structure of the working age population has significantly altered.

The methodology used internationally follows that of Shah and Burke (2001) used at the Centre for the Economics of Education (CEET) where a cohort component method was applied. The historical retirement rates were estimated using the 1996, 2001 and 2006 census data over the 2008/09 period.

Historical retirement demand was used as the basis to forecast net retirement rates over the medium term future (2009-19). The likely age composition (i.e. column shares) of occupational groups in future periods such as in 2011 and 2016 was emulated by studying the patterns that were observed in the past three Census of 1996, 2001 and 2006. Statistics New Zealand's Working Age Population (WAP) projections by age cohorts for the Medium Scenario (5M) was then used to derive the levels of implicit employment by various age cohorts for each of the occupational groups. These were then used to study the net outflows and hence the forecast net retirement rates for the forecast period.