National Labour Force Projections 2015-base

Reference period

This release contains 2015-base projections of the labour force usually living in New Zealand. These supersede the 2006-base projections released in August 2012. The new projections have the estimated resident population in the labour force at 30 June 2015 as a base, and cover the period 2016–68 at one-year intervals. The labour force projections are derived from the latest National population projections: 2014(base)–2068 (released 28 November 2014) by multiplying the projected population by the assumed labour force participation rates (LFPRs), by single year of age and sex. Detailed projection results, including projections for individual years and by single-year of age and sex, are available in NZ.Stat.

Changes since the previous 2006-base projections

Average hours worked

For the first time, these projections include projections of the average number of hours worked (or available for work) per week by those in the labour force. Projections of the total number of hours worked (or available for work) are derived by multiplying the projected labour force by the assumed average number of hours worked (or available for work), by single year of age and sex. These can then be divided by the projected labour force, in any given age-sex group, to give the average number of hours worked (or available for work) per week.

Review of assumptions

The derivation of the projections involves a review of all projection assumptions. The main changes from the previous 2006-base projections (August 2012 update) are:

• Slightly lower male LFPRs at most ages, especially at ages 15–16 and 78–84 years.
• Slightly higher female LFPRs at ages 20–55 years, and lower female LFPRs at most other ages, especially at ages 15–17 and 64–86 years.
• Slightly higher population projections in the long term. This reflects the combined impact of updated fertility, mortality, and migration assumptions.
• The median New Zealand population from the 2014-base projections is 4.59 million in 2015, 5.50 million in 2038, and 6.17 million in 2068. By comparison, the equivalent populations from the previous 2011-base projections were 4.54 million in 2015, 5.42 million in 2038, and 6.16 million in 2068.
• The median annual net migration gain is assumed to be 54,000 in 2015 and 33,000 in 2016. This compares with the 2011-base projections (medium variant) assumptions of 12,000 for both these years.
• The median period life expectancy at birth reaches 89.0 and 91.5 years for males and females, respectively, in 2068. This is higher than the corresponding figures of 88.9 and 91.3 years in the 2011-base projections (medium variant).

Projection assumptions

Projection assumptions are formulated after analysis of short-term and long-term historical trends, recent trends and patterns observed in other countries, and government policy.

Population projections

Labour force projections for 2015–68 are based on the population projections summarised in the release National population projections: 2014(base)–2068. In brief, these population projections assume:

• a base estimated resident population (ERP) of New Zealand of 4.51 million at 30 June 2014
• fertility rates varying throughout the projection period. The median period total fertility rate declining gradually from 1.95 births per woman in 2015 to 1.90 in 2030 and beyond.
• death rates varying throughout the projection period. The median assumption has male period life expectancy at birth increasing to 84.7 years in 2038 and 89.0 years in 2068. The corresponding female period life expectancy at birth increases to 88.0 years in 2038 and 91.5 years in 2068.
• migration varying throughout the projection period. The median net migration (arrivals less departures) decreases from 54,000 in 2015 to 33,000 in 2016, and to 12,000 in 2017 and beyond.

Labour force participation

Labour force participation rates (LFPRs) measure the proportion of the population in the labour force, either part-time or full-time. LFPRs differ significantly across age for both males and females.

Assumed LFPRs are formulated from analysis of trends in the Census of Population and Dwellings and the Household Labour Force Survey (HLFS). Although the same definition of labour force is used in the projections as in the census and HLFS, some important differences exist:

• The HLFS provides the official measure of the labour force using an interviewer-administered survey of about 15,000 households and 30,000 people each quarter. By comparison, the census provides a snapshot of the labour force (usually every five years).
• The HLFS measures labour force status over each quarter, while the census question refers to labour force status in the week before the census date.
• Unlike the HLFS, the census is not subject to sample error (although both data sources may contain non-sampling errors). As a result, the census can provide information at a more detailed demographic level (eg single year of age) than the HLFS.
• Non-response in the HLFS is minimised through the use of best survey practices. Because the census is self-administered, higher rates of item non-response occur.
• The HLFS generally excludes people in the armed forces and non-private dwellings (eg retirement homes, hospitals, prisons), while the census includes everyone who is in New Zealand on census night.

These differences explain why LFPRs, as well as numbers in the labour force, vary between census and HLFS. These differ again from the base for these labour force projections, which is the estimated resident population of New Zealand in the labour force at 30 June 2015.

Compared with the HLFS, the 2013 Census generally indicated higher LFPRs for males and females at ages 65+ years. The 2013 Census also indicated lower LFPRs for males at ages 25–54 years.

LFPR assumptions are formulated by single year of age and sex, and for each projection year including the base year. Important considerations in formulating LFPR assumptions are:

• comparability of LFPRs across age (eg consistency between adjacent ages)
• comparability of LFPRs across projection period (eg consistency between adjacent years)
• comparability of male and female LFPRs at each age and each projection year
• plausibility of LFPRs (eg 0 ≤ LFPRs ≤ 1).

The main features of the median LFPR assumptions between 2015 and 2068 are:

• increases in LFPRs for males and females aged 56+ years, especially for males aged 62–79 years and females aged 56–75 years, reflecting increased flexibility in the age of retirement (with no compulsory age of retirement), changing attitudes to retirement, and increasing life expectancy and well-being in the older ages.
• small increases in LFPRs for females aged 23–55 years, partly reflecting declines in completed family size and increases in childlessness.
• small decreases in LFPRs for males and females aged 15–17 years, partly reflecting increasing rates of participation in tertiary education.
• static LFPRs for males and females at other ages.

Future labour force participation trends are uncertain and depend on a range of factors.

• Changes in population composition and different trends in population subgroups (including ethnic groups).
• Trends in fertility including the timing and number of births.
• Trends in the patterns of education (especially tertiary education) and work, including the timing, duration, and proportion of time dedicated to those activities.
• Trends in the balance between paid work, unpaid work, family, and leisure activities at different ages.
• Changing macro-level conditions (eg global and national economic conditions, government policies) that affect the labour market and demand for labour.
• Trends in health and mortality, affecting well-being and life expectancy, especially at ages above 50 years.
• Changes in financial considerations, including eligibility for government superannuation, especially at ages above 60 years. Simulations of LFPRs are produced in two steps:

1. Simulations of average working life (AWL) – in this case the sum of LFPRs over ages 15–79 years, by sex – are produced using a simple random walk with drift model. Random errors are sampled from a normal distribution with a mean of zero and a standard deviation of 0.515 for males and 0.341 for females. The standard deviations are derived by fitting an autoregressive integrated moving average or ARIMA (0,1,0) model to annual AWL by sex for June years 1986–2015. The drift function shifts the median of the AWL simulations to follow the assumed median AWL.

2. The median and standard error for LFPRs by age-sex are formulated for each projection year from historical data. These LFPRs by age-sex are scaled to each AWL simulation. In this method LFPR simulations are correlated across age-sex (ie if LFPRs are high, they are high at all ages for both males and females), but vary randomly from year to year.

Average hours worked

Average hours worked (AHW) measures the extent to which the labour force is available for work. As with LFPRs, AHW differs significantly across age for both males and females. Assumed AHW are formulated from analysis of trends in the Census of Population and Dwellings and the Household Labour Force Survey (HLFS) based on average hours worked by those working. People in the labour force who are not working are assumed to be available for work to the same extent as those working.

Simulations of AHW by age-sex are produced in two steps:

1. Simulations of total hours workable (THW) – in this case the sum of AHW over ages 15–79 years, by sex – are produced using a simple random walk with drift model. Random errors are sampled from a normal distribution with a mean of zero and a standard deviation of 31.6 for males and 39.4 for females. The standard deviations are derived by fitting an autoregressive integrated moving average or ARIMA (0,1,0) model to annual THW by sex for June years 1987–2015. The drift function shifts the median of the THW simulations to follow the assumed median THW.

2. The median and standard error for AHW by age-sex are formulated for each projection year from historical data. These AHW by age-sex are scaled to each THW simulation.

In this method AHW simulations are correlated across age-sex (ie if AHW is high, it is high at all ages for both males and females), but vary randomly from year to year.