Summary

This report shows that it is feasible and it is likely to provide an improved measure for New Zealand data in OECD comparisons of earnings premiums. The report finds that the new estimates of annual earnings for people in the HLFS sample are around, on average, nine percent higher than survey week earnings multiplied by 52. The difference is a little higher for those with higher levels of education. The impact on earnings premiums is very small for levels below degree. The analysis shows that the New Zealand figures under the current method slightly under-estimate the earning premiums^[2] for people with a degree or higher qualification, as shown in the following table.

Results in more detail

Earnings premiums show the difference in earnings for people with different levels of educational attainment. It is one way that is used to show the benefits of education. For example, someone with a bachelors degree may, on average, earn 30 percent more than someone with a school qualification. The OECD compares and publishes these each year in its annual Education at a Glance (EAG) report. It uses the international standard classification of education (ISCED)^[4] to distinguish different levels of educational attainment.

Using Stats New Zealand’s Integrated Data Infrastructure (IDI), we were able to compare the annual taxable earnings from Inland Revenue (IR) data for people in the HLFS sample and measure the difference in earnings premiums between the two sources. We discuss how this was done in more detail at the end of the report.

Overall, there is a small difference in the premiums calculated through HLFS and tax data. However, the difference increases at higher levels of educational attainment. This shows that the current approach slightly under-estimates the earning premiums for people with a degree or higher qualification.

Figure 1: Difference between the EAG premiums between HLFS and IR in 2022^[5]

Note: the premiums shown here are relative to the ISCED 3 group (therefore ISCED 3 is 100 percent). The increase/decrease of premiums for all other groups are compared to the earnings of those with ISCED 3 qualification.

The highest difference in premiums is for men with education levels ISCED 6 and higher. The premiums for women, on the other hand, are slightly lower when based on the tax data compared to when based on the HLFS data. Those aged 35 to 44 have higher premiums across most ISCED levels, whereas those aged 55 to 64 have lower premiums when based on the tax data.

Figure 2: Difference in EAG premiums between HLFS and IR earnings by gender and ISCED levels

Figure 3: Difference in EAG premiums between HLFS and IR earnings by age group and ISCED levels

In 2022, the difference between the total earnings in the HLFS and the earnings for the same people in the tax data was around nine percent.^[6] Since 2016 this difference has stayed between five and 14 percent, with the highest (14 percent) in 2020 and the lowest (six percent) in 2018, with tax data reporting higher earnings each year.

This difference in earnings gets higher for higher-educated people i.e., those that have ISCED Levels 6 to 8 qualifications have higher annual earnings, on average, in the annual tax data compared to the HLFS. Those that have ISCED Level 6 qualification have nine percent higher earnings, and those that have ISCED Levels 7-8 qualifications have 11 percent higher earnings in the tax data compared to the HLFS. It also proves that the education to earnings ratio in New Zealand is slightly higher than what has been interpreted in the past. New Zealand has had the lowest returns on education compared to other OECD countries.^[7]

More detailed information of the difference in population and earnings is given below.

HLFS is a survey that collects information from a sample of population. Each individual from a sample is then assigned a weight based on various factors^[8] to reflect the total (targeted) population of New Zealand. These individuals are then matched to the IDI data where we can find additional information about them (matching them to other datasets). In this case, we match them to the IR data to capture their IR earnings. Earnings are then divided by 52 to get an estimate of IR weekly earnings to compare with the HLFS weekly earnings. Not every sample individual can be matched to the IDI as mentioned in the table above (with a match/without a match). Details on how these are matched (including the match rates) can be found in the methodology section at the end of this report.

Although everyone, on average, has higher earnings in the tax data compared to HLFS, we found that men have higher earnings than women in the tax data compared to the HLFS in 2022 throughout different ISCED levels apart from ISCED 3. Similarly, those that are aged 45 to 64, on average, have higher earnings in the tax data compared to the HLFS in 2022.

Figure 4: Difference in the proportion of HLFS and IR earnings by gender and ISCED levels

Figure 5: Difference in the proportion of HLFS and IR earnings by age groups

Over the years, the difference between the HLFS and tax earnings has stayed between six and 14 percent, on average, from 2018 to 2022, with the highest difference of 14 percent in 2020. The year 2020 was a bit out of trend for all different gender, age and ISCED groups.

The difference for each ISCED level between the two datasets has changed significantly from 2018 to 2022. In 2019, the tax earnings for those with ISCED Level 5 qualification were three percentage points lower than the earnings in the HLFS. It was the only year and the only ISCED group that had a lower tax earnings compared to the HLFS. The highest difference (20 percent) was in 2021 for those earners with ISCED Level 7-8 qualifications.

Figure 6: Trend in the difference between the proportion of HLFS and IR earnings by ISCED levels

Those earners who are aged 45 to 54 have been the most volatile over the last five years. While there was only a five percentage point difference between the two datasets in 2018, this rose to a 16 percentage point difference in 2020 and a 10 percentage point difference in 2022. The largest difference from year to year was for those aged 55 to 64, jumping from four percentage points difference in 2019 to 15 percentage points difference in 2020.

The trend looked similar for both men and women, with spikes in differences in 2020. The average difference between HLFS and tax earnings over the last five years (2018 to 2022) has been eight percent for women and 10 percent for men.

Figure 7: Trend in the difference between the proportion of HLFS and IR earnings by age group

Figure 8: Trend in the difference between the proportion of HLFS and IR earnings by gender

Data and Methodology

Every year the OECD asks its member countries for information on earnings by level of education, to show comparisons of the earnings benefits or premiums that higher levels of education can provide.

For New Zealand, the Household Labour Force survey (HLFS) captures all the variables that are needed for the data to be reported to the OECD, including educational attainment, earnings, age, gender, and whether working part-time, full-time. It is run frequently, and its sample is designed to provide representative estimates for the total population. Its limitation – in terms of the OECD reporting requirements – is that earnings are relative to one week in the year, rather than an annual total, as requested by the OECD. So, for the purposes of meeting these requirements, the weekly earnings are simply multiplied by 52 to calculate the annual total.

The purpose of this work was to explore whether we could compare the annual earnings for the people in the HLFS with their annual earnings from tax data, to see if this might represent a better estimate of annual earnings than multiplying the HLFS weekly earnings by 52. This in turn would provide a more accurate source for OECD’s comparisons of earnings premiums for different levels of education.

We use all 25 to 64-year-olds captured in each June quarter of the HLFS from 2018 to 2022 and match them to the tax data for the respective year in the IDI (using ‘snz_uid’ – a variable to identify unique individuals in the IDI environment) to calculate their total annual income for the particular tax/financial year.^[9] Some individuals captured in the HLFS will not have reported income in the IR, and we discuss how we have treated these below. For those aged 25 to 64 and employed – our target population – the match rate between the HLFS and the tax data for the last four years (2018 to 2022, which includes two pre-Covid years) is between 86 and 93 percent, with the lowest in the most recent year i.e., 2022.

Methodology

For those earners in the HLFS that do not have a match in the tax data in the IDI (for instance, 14 percent in June 2022), we tested two different methodologies for defining their income. For methodology 1, we simply replaced the earnings of those unmatched with their HLFS earnings (x 52). For methodology 2, we took the proportion of the earnings of tax data by HLFS earnings for the matched and applied the same proportion to the unmatched. This proportion has been calculated by dividing the matched into groups of gender (male and female), age (25-34, 35-44, 45-54, 55-64) and ISCED levels (0-2, 3, 4, 5, 6, 7-8). The characteristics of those unmatched and matched were similar by these categories.

Methodology 2 is conceptually more precise, but computationally more complex. There is almost zero difference in either method on the resulting published EAG premiums.

What types of tax data are counted as earnings?

Out of the different sources of income captured in the tax data, we only used the following to correctly match the IR reported income with the HLFS reported income:

• W&S – Wages and salaries.
• C00, C01, C02 – Company director/shareholder income from IR4S, receiving PAYE deducted income, receiving WHT deducted income.
• S00, S01, S02 – Sole trader income from IR3, receiving PAYE deducted income, receiving withholding tax deducted income.
• S03 – Rental income from IR4.
• P00, P01, P02 – Partnership-based income.

and exclude the following sources of income:

• BEN – Benefit payments from the Ministry of Social Development (MSD).
• CLM – Accident Compensation Corporation (ACC).
• PEN – Pension payments from MSD.
• PPL – Paid parental leave from MSD.
• STU – Student allowance from MSD.

HLFS reports on earnings on one week in the reference period of the survey (i.e., a week in the June quarter). Tax data reports annual earnings of the entire year or in this case each financial year (i.e., March to March). For our comparative analysis, tax annual earnings have been divided by 52 to get an estimate of weekly earnings as reported in the HLFS.

Classification of education

The OECD uses International Standard Classification of Education (ISCED) Levels for a comparable educational achievement indicator among all OECD countries. How these educational attainment levels compare among ISCED, EAG and New Zealand Qualification Framework (NZQF) is explained in the table below.

IDI confidentiality rules

The IDI confidentiality rules, on average, change the population and total income by less than one percentage point. This means that the Education at a Glance premiums do not change because of the confidentiality rules.

IDI disclaimer

These results are not official statistics. They have been created for research purposes from the Integrated Data Infrastructure (IDI) which is carefully managed by Stats NZ. The results are based in part on tax data supplied by Inland Revenue to Stats NZ under the Tax Administration Act 1994 for statistical purposes. Any discussion of data limitations or weaknesses is in the context of using the IDI for statistical purposes and is not related to the data’s ability to support Inland Revenue’s core operational requirements. Access to the data used in this study was provided by Stats NZ under conditions designed to give effect to the security and confidentiality provisions of the Data and Statistics Act 2022. The results presented in this study are the work of the author, not Stats NZ or individual data suppliers. For more information about the IDI please visit https://www.stats.govt.nz/integrated-data/.

Footnotes

1. Education at a Glance - OECD (2023). Chapter A, Indicator A4.
2. Education earnings premiums in Education at a Glance are calculated using International Standard Classification of Education (ISCED) 3 earnings as the base achievement.  Earnings for other ISCED levels are calculated relatively based on the earnings of ISCED 3.
3. The New Zealand Qualification Framework (NZQF).
4. international-standard-classification-of-education-isced-2011-en.pdf (unesco.org)
5. Refer to Table 6 below in the methodology section for a comparison between New Zealand Qualification Framework and ISCED levels of education.
6. IR earnings have been divided by 52 to get an estimate of IR weekly earnings as a comparable measure with HLFS weekly earnings.
7. Study on New Zealand’s education and earnings can be found here: Education and earnings, a New Zealand update | Education Counts. Comparative analysis between New Zealand and other OECD countries, including that of earnings and education, can be found here: How does New Zealand's education system compare? OECD's Education at a Glance 2023 | Education Counts
8. Details on how the selection of sample and assigning of weights to the sample can be found here: HLFS sources and methods: 2016 (stats.govt.nz)
9. We use financial year to account for all the self-employed during the reference period. Calendar year information in the tax data doesn’t necessarily include all self-employed as they are not required to file for a tax return until the end of the financial year.