<
>Keywords: Distributional Economics; Statistics;

 

 

Introduction [1]

 

One of the few useful sources of market income information is incomes declared for tax purposes. Even so it has limitations.

 

It is administrative data and so is sensitive to changes in statute and administrative policy. Thus the definition of income is that set by the law. In the New Zealand case the law generally omits capital gains. Coverage is affected by administrative practice. Low income recipients may not need to report their incomes (the government relying on PAYE to tax them).[2] The available data excludes trusts and companies.

 

Even so it is the best data we have. Robert Solow famously justified some statistical work he was doing, by citing the addicted gambler who knew ‘the casino wheel is crooked but it is the only one in town’.

 

In any case others have used the data.[3]This paper reports on and interprets their data in order to get a better understand what has been happening.

 

Data Sources

 

All the data used here ultimately comes from the Department of Inland Revenue. It is based on a sample of IR3 and IR5 tax returns for each of the 31 years from 1981 to 2011.[4]

 

The data consists of a taxable income band, the number of taxpayers in the band and their total taxable income.[5] The bands vary from year to year, but are narrow enough to enable the required calculations to be done with sufficient precision.

 

Defining the Top One Percent and Their Income Share

 

It was soon evident that the coverage of those reported in the data base has varied, reflecting the administrative decisions of who is required to file a return, together with the ability to reconstruct a return for those who are not required to file.

 

Taxpayers numbered in 1981 were 70.9 percent of the adult population over 15; in 2011 they were 98.2 percent.[6,7] The most important reason for the rise has been the inclusion of the PAYE record for individuals who did not file. It would be misleading to compare the one percent of tax reporters in 1981 with the one percent in 2011.

 

Instead, the one percent is defined here as the top one percent of ‘adults’ (over the age of 15)  irrespective of whether they appear in the IRD statistics.[8] That represents about 23,000 individuals in 1981 and 35,000 in 2011 with a corresponding (almost) linear growth between.

 

If taxpayer coverage varies over time, then so must the coverage of the aggregate incomes which they report. There is no uncontaminated benchmark series as for population. The best I could find comes from the Household Sector Account from the System of National Accounts of primary income receivable.[9] It does not match taxable income because ‘social security benefits in cash’ and ‘social assistance in cash’ are included in taxable income.[10]

 

The appendix table shows the resulting declared taxable income of the top 1 percent of adults, together with four other measures. as a share of aggregate income.

 

The Top Ten Percent

 

Exactly the same method was used to calculate the share of the top 10 percent. The difference between that share and the share of the one percent is the share of the next 9 percent which is also shown in the appendix table.

 

The top 10 percent amounted to 350,000 people in 2011. They reported more than $71,000 of taxable income in a year when the annualised average  wage was about $51,000. The closeness of the two reflects that there is a large tail of adults who are not full-time wage earners, including part-time workers, superannuitants, beneficiaries and other non-earners. (Only 22 percent of the adult population declared more than the annualised average wage.)

 

The Pareto Coefficient

 

Vilfredo Pareto famously proposed that upper incomes followed a power probability law which is today called the ‘Pareto distribution’. Its shape is characterised by a single parameter, the ‘Pareto Coefficient’, which he thought was universally near to 2. While practically top incomes roughly follow a Pareto distribution, the Pareto Coefficient itself is much more variable between countries and over time. In the case of early postwar New Zealand it was typically above 2.5, rising, and near 3.0 in the mid-1970s.[11]

 

The larger the Pareto Coefficient, the less unequal the distribution. (Thus the rising coefficient in the early part of the post-war era indicates that inequality was falling.) [12]

 

The Top 0.1 Percent

 

Overseas studies often report the share of the top 0.1 percent. Given the difficulties with the New Zealand data, any estimation can be treacherous. Statistically the tabulations are derived by sampling. Estimates using them will be less satisfactory where the numbers are small and the distribution highly skewed, as applies for the very top of the income distribution.

 

The issue of under-reporting for tax avoidance poses additional problems. Those with high incomes can avoid declaring offshore income for taxable purposes by using the residential rules and spending sufficient time out of New Zealand.[13] Trusts and income splitting may also be important.  (Such avoidance may also occur among those on lower incomes, although there is less incentive.)

 

As well as the reporting difficulties, the upper open interval is usually too big to enable a direct estimate. For instance in 2011 there were about the 3,500 in the top 0.1 percent of the population, but those in the upper open interval numbered 13,040.

 

Instead, the upper income tail is assumed to be Pareto distributed and the share of the top 0.1 percent is estimated from the share of the top one percent using the Pareto coefficient.

 

What Does the Top of the New Zealand Market Income Distribution Look Like?[14]

 

The Appendix Table provides statistics of the shares for the top income categories and the Pareto coefficients for each of the 31 (mainly March) years between 1981 and 2011 inclusive. Its interpretation is illustrated by consideration of the last (2011) year.

 

In 2011 the top one percent of individuals declared about 9.2 percent of the total market income of New Zealand. That means that their average market income was 9.2 times that of the average New Zealand adult.[15] The top 10 percent declared 35.4 percent share of the total income, so they averaged 3.5 that of the average New Zealander. The top 0.1 percent declared that they received (after avoidance) 22 times what the average New Zealander received.

 

How do these shares compare internationally? Such comparisons are very difficult because they may cover different populations (notably taxpayers rather than adults) and income aggregates (declared taxable income rather than the total). For the record, there is data for 18 OECD countries; their average for share of the top one percent comes to 9.8 percent, so the New Zealand share appears to be near the average of the reported OECD shares.[16]

 

The following may be useful benchmarks for the 2011 tax year:

10 percent of adults had incomes above about $70,000

1 percent of adults had incomes above about $165,000

0.1 percent of adults had incomes above about $420,000

Annualised average Wages = $51,000

 

The Pareto coefficient was 2.6; it cannot be readily interpreted for these purposes. However that the average of 17 OECD countries in 2005 came to 2.1 so New Zealand was probably less unequal on this measure, although it may be easier for rich New Zealanders to move offshore or into capital gains in order to escape the full impact of the New Zealand taxation regime.[17]

 

Inequality at the Top over Time (See The Appendix Table and Figure)

 

The four indicators change over the 31 years. There is probably not a lot to be gained from a year-to-year analysis because of sampling variability. However there are distinctive trends. Most fundamentally, the share of the top groups is higher at the end of the 31 years than at the beginning, while the Pareto Coefficient is correspondingly lower.[18]

 

Thus the top one percent had a share of 6.0 percent in 1981 and 9.2 percent in 2011. (The share of the top 0.1 percent rose from 1.2 percent to 2.2 percent in the same period. The top 10 percent from 30.1 percent to 35.4 percent.)

 

The change over time is not smooth; for all measures most of the changes occur in the early 1990s. The share of the top 0.1 percent increases by about 1.5 percentage points between 1989 and 1993; the share of the top one percent leaps from roughly 6 percent to roughly 10 percent between 1989 and 1993 (Since this includes the 0.1 percent, it means that the remaining 0.9 percent get about an additional 2.5 percentage points between them). The share of the top 10 percent shows a leap of about 10 percentage points in the same period (so the next 9 percent have a lift of about 6 percentage points). Similarly the Pareto Coefficient hovers around 3 in the 1980s (a level similar to the late 1970s) and then falls to 2.3 in 1993.[19]

 

After 1993 there seems to a mildly rising trend peaking at the turn of the century, and then falling. The lessening of inequality seems to precede the arrival of the Global Financial Crisis. Or it could be argued that the trends after 1993 are broadly flat.

 

Why the Increase?

 

Before trying to answer the question as to why there has been an increase in inequality at the top, three points need to be cleared away:

– this is a personal income distribution, it is not the same as the household income distribution, which much of the New Zealand income distribution debate has focussed on, and does not map easily to it.[20]

– this is the market distribution and, subject to the discussion on dividend imputation below, does not take into account income tax changes which, generally favouring the rich, have increased their share of after-tax incomes even further. (Almost all the discussion on the household income distribution is after-tax and with cash social assistance.)

– the dramatic change occurs in a very short period – not earlier than 1989; not later than 1993. There is little evidence in the data of a long term trend (nor – perhaps more surprisingly – of a business cycle).

 

What might have been the causes of the widening of the income distribution at the top? The main reasons seem to be two fold.

1. The Impact of the Dividend Imputation System

 

Until 1989 it was said that corporate dividends were ‘doubled taxed’. First corporate profits paid corporation tax and then, dividends paid from the tax-paid profits were treated as taxable income of the shareholder. Of course in practice things were much more complicated for there was considerable tax avoidance.

 

From 1989 there has been a dividend imputation system in which a shareholder receiving a dividend from a company is entitled to an ‘imputation credit’, which represents tax paid by the company, and is used to reduce or eliminate the shareholder’s income tax liability.[21] In effect corporation tax becomes a withholding tax for shareholders’ dividends.

 

The (intended) effect was to encourage the distribution of corporate earnings with the consequence that shareholders were less dependent on capital gains and other artifices (such as returning to shareholder their capital investment) to access their share of corporate profits. The alternatives are explicitly designed to avoid reporting personal income; a dividend imputation system does not have the same disincentive (while giving the shareholder easier access to corporate profits). In effect it puts some previously unreported income (including capital gains) on the IRD books.

 

This altered the way that dividends are recorded in the tabulation. Thus $100 of corporate profits which were taxed at, say, 33 percent and fully paid out were recorded as $67 before imputation but as $100 after the new regime was introduced. This is in addition to the encouragement to corporations to pay more dividends which get reported for income tax purposes.[22]

 

In the 1988 year, (natural) persons declared $152m of net dividends. By 1993, after the imputation was introduced and had settled in, they were declaring $691m in net dividends which grossed up (i.e. with $252m of the corporation tax paid on them added back) came to $943m. That meant that reported dividends jump from about 0.4 percent of estimated market income to 2.1 percent.[23]

 

We are unable to attribute all this increase to those with incomes in the top 10 percent since some of the other 90 percent receive some dividends. However it would appear that as a result the top one percent declared an income boost of about 1 percentage point share of total income.

 

The share of declared dividends in total income continue to rise. By 2011 they probably amounted to 3.0 percent, up about another 1 percentage point on 1993.[24] However there was an increase after 2000 in income routed through trusts. It is possible that the turn down in the share of the top 1 percent in the following decade reflects this. In which case it is possible that, had there been no acceleration in the amounts distributed through trusts in the decade, the share of the top 1 percent would have been a percentage point (or even two) higher and there would have been no fall-off after 2000.

 

Strangely there is not evidence that this later increase in dividends increased the top share of incomes.[25] In any case it only explains a part of what was going on in the early 1990s.

 

2. Rewards for managers rose relative to rewards for labour

 

It would seem that the rewards for managers (and higher professionals) rose relative to the rewards for ordinary labour. This can be illustrated by considering the ratio between the salary of the Secretary of the Treasury and average earnings.[26] In 1981 it was about 5.6 times of the average wage but in 2011 it was almost double that at about 11.1 times. The Secretary’s remuneration pattern is little different for many senior civil servants, and was justified by a similar shift in top management rates in the private sector.

 

Exactly why this happened is complicated, but how it happened is not. The 1988 State Sector Act abandoned the rigid relativities that existed in the public service, enabling higher relative remuneration to the top civil servants, while most civil servants were experiencing restricted increases (or declines). Probably the same thing was happening in the private sector. (An important factor may have been the globalisation of the market for management and higher professional.)

 

It will be noted this change just preceded the time when the share of the one percent increased markedly.

 

Unfortunately however, we cannot estimate its total magnitude to assess to what degree that explains the rest of the upshift.[27]

 

Conclusion

 

The purpose of this paper has been to use IRD data to track changes in top incomes. Although some calibration is necessary and there has to be caution over omissions it has been possible to track from the tax year 1981 to tax year 2011 (the latest for which IRD data is available).

 

The basic conclusion is that there has been an increase in inequality, in that top incomes have increased faster than incomes as a whole. However most of this increase occurred in the 1989 to 1991 period probably because of an increase in remuneration margins for management and the introduction of a dividend imputation system.

 

There is some evidence that the share of top market incomes gradually rose after 1993 through to about 2000 (albeit interpretation of the data is complicated by some tax avoidance) and then as gradually fell away. The indicators suggest that on the measures of top incomes, inequality was lower in 2011 than in 1993 but higher than in 1988.[28]

 

Calibration difficulties make international comparisons difficult, so we must be cautious about ranking New Zealand’s top income inequality with economies elsewhere. However there is no evidence of a surge in inequality in the New Zealand data in the first decade of the twenty first century, as has occurred in the UK and the US, probably because New Zealand does not have as sophisticated financial sectors as they have. If anything, the share of those with top incomes seems to have fallen slightly in that period, perhaps as a consequence of the Global Financial Crisis.

 

Endnotes

[1] I am grateful for comments on earlier drafts by Norman Gemmell, Bill Rosenberg and Sandra Watson.

[2] For instance, in the 1981 tabulation excludes taxpayers who earned less than $11,500 per annum and were not required to file tax returns.

[3] The World Top Incomes Database http://topincomes.g-mond.parisschoolofeconomics.eu/

[4] The data for 2002 to 2011 is on their website. I asked the Department to supply me with the data file released in 2004 to Andrew Leith and Tony Atkinson from 1981 to 2003, which they did promptly. I am grateful for assistance from Sandra Watson, Manager, Forecasting and Analysis, Policy and Strategy, Inland Revenue.

http://www.nuffield.ox.ac.uk/users/atkinson/AtkinsonLEIGH_NewZealand08.pdf, page 46.

The two sets of series overlap by two years, a comparison of which confirms they are the same series.

[5] The data covers natural persons and excludes trusts and corporations.

[6] Tax years. For most taxpayers these are year ending in March

[7] The 98 percent includes children and those earning wages for part of a year then moving overseas.

[8] Under 15s may appear in the list, but there will be very few of them in the top one percent.

[9] It is based on Table 2.11 of national disposal income, The series was kindly supplied by Jeff Cope of SNZ.

[10] We are assuming that the top ten percent receive little of these transfers. Some superannuitants do.

[11] B. H. Easton (1983) Income Distribution in New Zealand, p.181.

[12] The coefficients were estimated by using the simple relationship between the mean and the bottom income of a Pareto distribution.

[13] See New Zealand tax residence. Who is a New Zealand resident for tax purposes?

http://www.ird.govt.nz/resources/a/3/a35bbb804bbe588fbc1efcbc87554a30/ir292.pdf

The general rule is summarised by

‘You’re a New Zealand tax resident if:

• you’re in New Zealand for more than 183 days in any 12-month period, or

• you have an “enduring relationship” with New Zealand, or

• you’re away from New Zealand in the service of the New Zealand government.’

People who are not New Zealand tax residents (non-residents) are liable for New Zealand tax only on their New Zealand-sourced income.

[14] As declared for tax purposes.

[15] On the basis of the handful of countries for which there are estimates the inclusion of capital gains might add about 1 percentage point to the top 1 percent’s income share, although having no capital gains tax to speak of is likely to incentivise accruing income in that form..

[16] http://topincomes.g-mond.parisschoolofeconomics.eu/#Home: The denominators may not be consistent.

[17] A B Atkinson, T. Piketty & E. Saez (2010?) Top Incomes in the Long Run of History

http://piketty.pse.ens.fr/fichiers/public/AtkinsonPikettySaezOUP09summary.pdf, Table 13.1.

[18] There is a marked blip in the March 2000 year. The Labour Government had been elected in 1999 on a promise to raise the top income tax rate from April 2000. Many of the rich arranged their income flows to pull income forward into the lower tax year. However the income shares fall in the following few years and a moving average over the period shows only the most marginal peaking in 2000. See I. Claus, J. Creedy & J. Teng (2012) The Elasticity of Taxable Income in New Zealand Ne W Zealand Treasury Working Paper 12/03.

[19] Perhaps surprisingly, there is no evidence of the impact of the 1987 sharemarket boom and bust in the data.

[20] B. H. Easton (2013) Economic Inequality in New Zealand: A User’s Guide.

http://www.eastonbh.ac.nz/2013/12/economic-inequality-in-new-zealand-a-users-guide/

[21] Dividend imputation is a corporate tax system in which some or all of the tax paid by a company may be attributed, or imputed, to the shareholders by way of a tax credit to reduce the income tax payable on a distribution. In comparison to the classical system, it reduces or eliminates the tax disadvantages of distributing dividends to shareholders by only requiring them to pay the difference between the corporate rate and their marginal rate. http://en.wikipedia.org/wiki/Dividend_imputation.

[22] We could have a vigorous argument at this point as to what extent the declared increase reflected a genuine increase in market incomes and to what extent it was the result of closing a tax avoidance loophole and an accounting convention of reporting them grossed up, the effect of which was to reduce their tax liability. The second interpretation would suggest the change belongs more to a change in after-tax incomes. We leave the debate to another venue. Our purpose is of measurement.

[23] The income indicator measures dividends grossed up.

[24] Additionally 2011 may have been a low year because of changes in tax treatment.

[25] Of course some dividends are flowing through trusts.

[26] The Secretary of the Treasury was paid $61,953 in 1981 and about $565,000 in 2011.

[27] Tim Hunter of The Dominion Post kindly made available a data base he has been collecting of the pay of chief executives of 34 listed companies. It suggests that their remuneration rates increased by about 10.7 percent p.a. in the 2009 to 2012 period; the data base shows an increase of market income per adult of 1.3 percent p.a. over the same period.

[28] However post 2000 is complicated by the 39c rate avoidance -the more fungible income moved into other entities, and was no longer declared by natural persons.