Top incomes and the gender divide

Abstract

In the recent research on top incomes, there has been little discussion of gender. A great deal is known about gender differentials in earnings, but how far does this carry over to total incomes? This paper investigates the gender divide at the top of the income distribution using tax record data for eight countries with individual taxation. It shows that women are strongly under-represented at the top of the distribution. Although the presence of women at the top has increased over time, the rise becomes smaller at the very top. The income gradient by gender has become more marked. The paper also shows that income composition differs by gender and exhibits significant changes over time, underlining the fact that it is not sufficient to look only at earned income.

Appendices

Appendix A: Detailed information on sources

Australia

Data source::

Australian Taxation Office (ATO)

Data format::

customised tables for 2001/02 to 2014/15 and tabulated data, 200 income ranges per year.

Population coverage::

whole population of individual taxpayers with positive net tax.

Definition of total income::

figures are based on total individual income before tax (adjusted for deferred losses) excluding capital gains and imputation credits. Due to data constraints, the Pareto analysis is conducted on total income excluding capital gains only. Figures 7 and 8 in Appendix B reports the share of women in top incomes in Australia based on total individual income excluding capital gains only and the share of women in top incomes including capital gains and imputation credits.

Tax year::

ends 30 June; labelled with the year of starting month.

Other information::

more than 70% of tax returns are filled with the help of a tax agent. The proportion is increasing with individual income, reaching 85% or more for the top 5% (top 5% figures based on data from the 1-2% micro samples).

Canada

Data source::

Statistics Canada

Data format::

tabulated data from official publications from 1942 to 1990. Figures computed by Statistics Canada based on the Longitudinal Administrative Database (LAD) from 1993 to 2015. LAD figures were downloaded from the Statistics Canada website, see Statistics Canada (2017b).

LAD population coverage::

the LAD is a 20% random sample of Canadian tax filers.

Definition of total income in LAD::

total individual income before tax (the market income plus government transfers and refundable tax credits) not including capital gains.

Tax year::

calendar year.

Other information::

figures on top percentiles from the LAD are computed with respect to the population of tax filers and not the population aged 20 and over (reference population used in the top income literature for Canada). The population of tax filers has increased over time. In the mid-80s, its size was about 85% of the size of the population aged 20 and over. From 1993 onwards, the number of tax filers represents more than 95% of the population aged 20 and over. This is the reason why we opted to consider Statistics Canada LAD data from 1993 onwards. Figures going back to 1982 are available on the Statistics Canada website. A chart that highlights how the historical series (based on tabulated data) and the LAD data series compare over time is reported in Fig. 12 in Appendix B.

Denmark

Data source::

Statistics Denmark

Data format::

micro data from 1980 to 2013.

Population coverage::

whole population of individual taxpayers, aged 15 and above.

Definition of total income::

total individual income before tax (market income + government transfers) excluding dividends and capital gains. Dividends and capital gains could only be identified jointly. The treatment and coverage of capital gains changed over the years but is included the data from the mid-2000s. We are grateful to Jakob Søgaard for pointing this out. As a result, the series that excludes capital gains is likely to provide a more reliable picture of the evolution of total income over 1980 to 2013. The series with capital gains should be more accurate regarding the level of total income at the end of the period.

Tax year::

calendar year.

Italy

Data source::

MEF Department of Finance/ Ministero delle Finanze (MEF)

Data format::

tabulated data from 1999 to 2015, 33–34 income ranges per year.

Population coverage::

whole population of individual taxpayers.

Definition of total income::

total individual income before tax.

Income sources not included or covered::

Dividends and capital gains are not covered unless received from qualified participation in a listed firm. Interest income not included. Dividends distributed by non-listed companies are included via the fiscal declaration of firms.

Top coding or grouping::

when frequency is less than 4 units, data are omitted for privacy reasons.

Tax year::

calendar year.

Other information::

in 2012 imputed rents pertaining to the house where the owner lives (redditi fondiari derivanti da beni non locati) are excluded from total income before taxation. From 2011, there is the option to pay a withholding tax on income derived from renting a house (rather than the progressive personal income tax). Starting from 2009 there is a 10% withholding tax on wages coming from extra-time. To subtract this income component from progressive taxation, there is a wage limit of 35,000 increasing to 40,000 euros.

New Zealand

Data source::

New Zealand Inland Revenue

Data format::

published tabulated data for most years between 1953 and 1989 (see Atkinson and Leigh 2007b) and tabulated data provided by the Inland Revenue from 1980 to 2015, with 52 to 172 income ranges per year. A chart that highlights how the historical and recent data series compare over time is reported in Fig. 13 in Appendix B.

Population coverage::

the recent tabulated data is based on a random sample of individual taxpayers, scaled up to population estimates. The sample is 2% of wage and salary earners and 10% of IR3 filers. Data from 1981 to 1993 include people who filed income tax returns. People were not required to file tax returns if their income was below a specified threshold unless they had income sources from which withholding taxes had not been deducted. The data from 1994 onwards also include non-filers with PAYE income. The income data for such non-filers is sourced from employer-records, including taxable transfers.

Definition of total income in recent tabulated data::

total taxable individual income. Note that there are very few allowances available.

Tax year::

ends 31 March – labelled with year of starting month.

Other information::

gender identification in the recent tabulated data is based on the person’s title, as a proxy variable (random allocation of people with titles like Dr, Reverend, etc.). Structural break in 2000 when the top personal income tax rate was increased and the trust rate was not, resulting in many individuals channelling their income through trusts. The two rates were realigned in 2012, although ownership structures are likely to remain.

Norway

Data source::

Statistics Norway

Data format::

micro data from 1993 to 2013. Files of taxpayers linked to population registry.

Population coverage::

entire population.

Definition of total income::

total individual income before tax (includes some capital gains), see Aaberge et al. (2016) and Aaberge and Atkinson (2010). Note that due to data constraints the definition of total income used in the time series is not exactly the same as the definition of total income used in the income composition analysis for 2013. There was also a change in the definition of income in 2005, which may affect the time trends observed in the data.

Income sources not included or covered::

income from owner-occupied houses and non-taxable capital gains.

Tax year::

calendar year.

Other information::

tax changes on dividends in 2001 and 2006 affected the income reporting behaviour of capital owners and trends in investment income during the period from 2000 to 2013, see Aaberge et al. (2013).

Spain

Data format::

data for 1999–2001 comes from the “Panel de Declarantes de IRPF 1999–2009” which is a stratified random sample (panel) covering 2% of taxpayers. Data for 2002–2013 comes from the “Muestra de Declarantes de IRPF”, which is a stratified random sample that includes 6 to 8% of individual taxpayers. From 2002 onwards, the Muestra is used as it performs better when reproducing tabulation-based top shares. The Panel loses precision with time. Panel sample size about 390,000 to 425,000; Muestra sample size about 907,000 to 1,352,000.

Definition of total income::

total individual income before tax, excluding capital gains.

Income sources not included or covered::

Capital gains are covered at varying degrees.

Tax year::

calendar year.

Other information::

married couples can choose to file their tax returns jointly. When there is joint filing, it is not possible to distinguish the incomes of the members of the couple separately. In those cases, the micro-data base identifies only the gender of the individual with highest incomes in the couple. When both members of the couple have non-zero incomes, the joint filing is clearly disadvantageous, the more so for relatively high incomes. Around 20% of the tax returns in the top 10% and top 1% of the income distribution are joint. We have considered that all joint files at the top of the distribution are individual (so one of the members in the couple has zero or very little income), and attributed them to the member of the couple with highest incomes, as done in the micro-data base. Throughout the period, 90-95% of joint tax returns in the top 1per cent group (and 80–90% in the top 10% group) correspond to men. It should also be noted that a small fraction of joint files corresponds to sole-parent families.

United Kingdom

Data source::

HMRC, Survey of Personal Incomes (SPI)

Data format, year and sample size::

data for 1995/96 to 2010/11 and for 2013/14 to 2014/15 are drawn from micro dataset of the Survey of Personal Incomes (no data were released for 2008/9). For details on data sources see entries under ‘HM Revenue and Customs’ and ‘Inland Revenue’ in the reference list. The figures for 2011/12, 2012/13 and 2015/16 are based on interpolations of the published tabulations of the Survey of Personal Incomes. Where the tabulated data for the UK are used, the relevant top percentile is calculated applying a Pareto interpolation to the whole distribution (male and female); the overall percentile is then used to calculate the number of women by applying a Pareto interpolation to the distribution for women alone.

Sample size: 57,400 to 729,300 observations per year in the micro data, 13–14 income ranges for the tabulated data. Note when using the micro datasets, data points based on less than 20 (unweighted) individual observations are not considered in the analysis.

Population coverage::

representative sample collected from 3 different sources: (1) the National Insurance and PAYE Service, (2) the Computerised Environment for Self-Assessment system which covers people with self-employment, or people who receive rental or untaxed investment income, and (3) the Claims system, which covers people who are not generally taxpayers but who have had too much tax deducted at source and have made a claim for its return.

Definition of total income::

total individual income before tax.

Income sources not included or covered::

some social security benefits and income from some tax efficient savings vehicles that are not taxed. Capital Gains arising from the disposal of assets are subject to Capital Gains Tax (CGT) and are not treated as income for income tax purposes, so gains from the disposal of assets are not included in the SPI. Some investment income is imputed.

Top coding or grouping::

in the micro datasets, the richest individuals are grouped into composite records.

Tax year::

ends 5th April – labelled with the year of the starting month.

Other::

independent taxation was introduced in the UK in 1990.

“In March 2009, it was announced by the Labour Government that the top rate was to be raised from 40 to 50% with effect from April 2010, and this led to “considerable fore-stalling” of income in 2009-10 (Atkinson and Ooms 2015). In March 2012, it was announced by the Conservative Government that the top rate was to be reduced to 45% with effect from April 2013, which again provided an incentive for income to be moved between tax years, in that case from 2012-13 to 2013-14.” Atkinson and Ooms (2015), p. 1.

Appendix B: Additional figures

Fig. 7

Australia, share of women in top income groups, total income excl. cap. gains only, 2000–2012

Fig. 8

Australia, share of women in top income groups, total income incl. capital gains and imputation credits, 2000–2012

Fig. 9

Canada, share of women in top income groups, total income incl. capital gains, 1993–2015

Fig. 10

Denmark, share of women in top income groups, total income incl. dividends and capital gains, 1980–2013

Fig. 11

Spain, share of women in top income groups, total income including capital gains, 1999–2013

Fig. 12

Canada, comparison with LAD (dashed lines) 1980 to 1995. Data for the LAD series starting in 1982 are included for comparison. We are considering LAD data from 1993 onwards

Fig. 13

New Zealand, comparison with historical data (solid lines) 1975 to 1992