NZ inequality statistics: Some of the research

Donal over at Economics New Zealand posted up some OECD figures that indicate that the Gini coefficient over the OECD was the same in 2010 as it was in the mid-1990s, and that it is actually lower in New Zealand.

As I have noted earlier, I am going to start writing about inequality on the blog.  So I have been spending a little bit of time reading about it!

Given this, I’ve realised we can take this analysis a step further.  Bryan Perry from MSD discussed the Gini coefficient, and other indicators, in his introduction for the inequality conference in July.  I wasn’t there – but I know the document is here, and I know Figure D.17 (third page of the pdf) has a graph of the Gini coefficient through time, and a trend line through it.

A couple of things should stand out when we look at this:

  1. The Gini coefficient has more been “flat” rather than “falling” since the mid-1990s if we look at the trend – the drop the OECD recorded looks like it may have been from comparing direct points, which are volatile
  2. When people complain about the large increase in the Gini coefficient they are not talking about the mid-1990s to today – they are talking about the reform period.  This figure shows that there was a very sharp increase in the Gini coefficient between about 1987 and 1992.

So unlike other countries, the complaints are NOT about a creeping increase in inequality through time – but about the level shift in inequality that New Zealand experienced following the reforms.  Ultimately, there is a view by these groups that the “equity-efficiency trade-off” New Zealand decided to make at that point wasn’t the right or just one.

Now I am not sure how we are even supposed to evaluate that claim without thinking about why, and how, inequality has changed.  To give some flavour for this, I’ll comment on a few of the New Zealand specific research papers we have had about this change – if you know any other similar work, flick me a line in the comments 😉

It was the reforms, and the reforms were for the rich – this is obvious, why are you asking

Stop stop, give me a chance here.  The reform period involved two direct channels that can impact upon measures of income inequality. [Note:  I am going to ignore changes to the way the HEIS/HES was measured during this period, as this is largely inconsequential]

The reforms were massive, they had a significant impact on the economy and society, and they were also directly followed by a massive recession – New Zealand’s own version of the Global Financial Crisis.

Now the collapse wasn’t so much a policy choice, although our central bank was determined to get inflation down and our central government was tightening its belt we also had:  the 1987 stockmarket crash, the collapse of the non-residential building industry, terms of trade collapse (oil price spike during Iraq war, falling prices for export commodities), the failure of BNZ and a corresponding financial crisis, the threatened credit downgrade of government debt (given the large stock of debt we had accumulated due to ‘Think Big’), then in 1991/92 we experienced a drought and electricity rationing.

I am not saying that the reforms were handled perfectly, and that they didn’t leave us fragile to shocks – but a lot of things went against New Zealand during this period.

Now, once the New Zealand economy had recovered, inequality measures had stayed higher than they were previously – in this case we may directly want to say the reforms increased inequality.  However, this begs the question “how”.  There are a two broad ways reforms may lead to a change in the observed income distribution:

  1. By changing tax rates, labour laws, or competition reform it changes the distribution of income among groups.
  2. By changing tax rates and the tax structure it changes the reporting of money income by groups.

Note that as a starting point I’m ignoring the absolute level of income, and the complicating impact of relative price shifts – so if there was an equity-efficiency trade-off, and if the reforms increased underlying incomes, then looking solely at the shape of the distribution would exaggerate the negative impact on the poor.  For example, the reforms drove down the price of basic manufactured products, something that was especially useful for poorer families – but this type of factor will be put to the side.

The first channel is the obvious one, and depending on our definitions of need among groups etc changes in this distribution may be seen as good or bad.

The second channel is more interesting.  For example, in 1985 fringe benefit taxation was introduced in New Zealand.  Suddenly, people who had been remunerated with goods and services (as a way to avoid tax) determined with their employer that they would rather just have cash.  If this type of practice was more common higher up the income distribution, this would lead to an increase in the Gini coefficient and the dispersion of money income – even though individual and households claim on resources has not significantly changed!  It is this second reason that may imply that the pre-1988 income data is largely incomparable with the post-1987 data – an important point to keep in mind.

Is there anything actually written about this stuff?

There are a few papers out there that try to look at the change in the income distribution (either with reference to a Gini coefficient or the full distribution itself) during this period.  Hyslop and Mare (2004) is the closest to what we are actually after here (along with Des O’Dea) – given that it’s goal is to understand why the shape of the entire income distribution shifted in the 1983/86-1995/98 period.

It is difficult to make strong conclusions in this type of literature, a point the paper is very keen to make clear – this type of work doesn’t allow us to definitively say anything, but helps use data to help shine light on what is going on, and update our priors appropriately.

Hyslop and Mare point out that, as well as being “due to the reform”, changes in the distribution of income may be due to:

  1. Changes in household structure across the population
  2. Changes in the socio-demographic characteristics of the population
  3. Shifts in employment outcomes within households
  4. And changes in the economic returns to attributes

Note that there was also a significant change in national superannuation (increasing payments for single-person households, cutting them for multi-person households) – something they also took into consideration.  Also note that these “changes” cannot be separated from the reform – if the reforms were to directly change any of these attributes, then that is indeed a result of the reform.

This paper, and the Des O’Dea paper, find that the changing households structure and socio-demographic characteristics elements account for a significant chunk of the ultimate change in the income distribution.

A follow up paper by Hyslop and Yahanpath (2005) focused on the 1998-2004 period, and instead looked specifically at working age individuals (and with a different data set – the HLFS income supplement).  Here they found that income growth was relatively evenly spread – but while high income individuals were gaining from wage growth, lower income individuals were seeing hours rise.  These leads in nicely to the arrival of Working for Families, and the high employment rates and low labour productivity we have been seeing.

So what

Good point, I’m going on a tangent.  So those above characteristics may be “valued” by our imagined ‘altruistic social planner’ in different ways.  For example, if incomes are becoming more disperse because the population is ‘older’ (implying that we are seeing people who invested in human capital now getting rewarded, and those that didn’t are not) then that may not be seen as a concern.  However, if we believed this was because the reforms “tore apart the family unit” we may take it as a bad thing.

To evaluate the reforms, we actually need to look at the reforms rather than aggregate indices!  However, there is no harm in trying to evaluate aggregate indices.  In this vein we have seen:

  1. There are another three papers by Podder and Chatterjee.  The first is (2000) and discusses changes in earned, ‘unearned’, and government cash benefit income between 1983/84-1995/96, and their corresponding impact on measures of the Gini coefficient.  The second (2002) and the third (2007 – with Dalziel and Daunfeldt) discuss the changes in income levels and distribution during this time in relation to ‘generalized Lorenz dominance’ – essentially asking if, for a given population, which realised income distribution would be preferred.
  2. Also Treasury has been interested in looking at “final incomes” which includes payments to, and benefits from, government.  Crawford and Johnson (2004) discuss this as does Aziz et al (2012).

While these are all cool papers that tease interesting facts out of the data, none of them can give us a comprehensive view on the reform – and none of them truly intend to of course.

What they do show is that there are a variety of reasons why an aggregate index like the Gini coefficient may have changed in New Zealand – and we need a clear idea about these reasons before we can discuss fairness, and before we can really think of policy.

  • VMC

    I would be interested in some thoughts on whether the Gini measure is actually fit for purpose – in the sort of discussion you are having. One possibility – actual incomes might have got more equal but if costs increased disproportionately on the poor, or govt subsides for the rich increased (for example) then inequality would increase where it mattered, if not in incomes.

    • The Gini index is useful for looking at inequality in a given distribution at a point in time – judging that, and figuring out what it means requires a lot of additional pieces of data and assumptions.

  • Donal Curtin

    My post about New Zealand’s “reduction” in income inequality was (I confess) slightly tongue in cheek – the key period for changes in New Zealand inequality (as you note) was in the decade leading up to 1995, rather than the period since (which was chosen by the OECD). Even so, I was somewhat surprised that aggregate OECD inequality didn’t appear to rise since 1995, and either stabilised or fell slightly in NZ, depending on how you look at it, given that for most of the period there was bloat at the top end of the income distribution. Incidentally one of the points the OECD noted was that “Eastern European and Nordic countries display the lowest income inequalities. However it was in the latter that income inequalities rose the most in the past 15 years, a trend which has been observed though to a smaller extent in many other European countries…”. It seems to me that the Nordic countries were also likely revisiting the calibration of that “equity-efficiency trade-off” you mention, and concluding that redistribution on a Nordic scale was having unacceptably adverse effects on incentives at both ends of the income distribution

    • Heya,

      Indeed, I picked it was a bit tongue in cheek – but it was a good lauching pad for a post I was writing anyway! I love it when that sort of stuff happens!

      The issue of inequality is fraught with ambiguities, and without considering the equity-efficiency trade-off societies care about we can’t say too much. I am a touch surprised to see how much the issue has been jumped on around the world and in NZ. Keeps economists busy though!