The trouble isn’t so much the things we don’t know, as the old aphorism goes, but rather the things we know that aren’t so. Simply not knowing things can often be fixed. Or, if it cannot be fixed, at least we know that we don’t know.

But received public wisdom seems to know a few things that aren’t so. And that seems especially true about income inequality. Everyone seems to know that income inequality is increasing – despite it just not seeming to be so.

Every year, the Ministry for Social Development releases its compendium of household income statistics. That careful work by the Ministry’s Bryan Perry explains the latest snapshot provided by the Household Economic Survey and shows how things have changed over time. Income inequality rose from the mid-1980s through early 1990s but has been generally flat since then. The Gini score is one of many inequality measures that show roughly the same pattern.

Figure J.13 Gini scores (x100) for market and disposable household income, 1985 to 2016 (18-65 yrs)

Perry’s work also shows that income growth has been reasonably broadly shared. These kinds of distributional analyses line up households from lowest income to highest income, then measure the earnings of the household at the top of each decile band. A household at the top of the second decile earns more than 20 percent of households, and less than the other 80 percent. Over the past decade, while there has been slightly higher income growth in the higher income bands than the lower income bands, real incomes are up by at least ten percent across-the-board.

 Figure D.2: Real equivalised household incomes (BHC): changes for top of deciles, HES 2008 to HES 2017

While Perry’s work sounds decidedly rosy by comparison to the more typical stories about lower-income stagnation, it understates real improvements in earnings in the lower half of the income distribution.

The Household Economic Surveys provide an annual snapshot of earnings but do not track how any particular individual or household’s income has changed over time. If the same households were locked into the same deciles every year, then the slightly higher income growth in the eighth and ninth deciles could compound over time into larger disparities in earnings.

But that is hardly all that is going on. Most people begin work at relatively low wages before earning more as they gain qualifications and experience. People move through the income deciles as they shift from education into work, and later into retirement. And a lot of people have highly variable earnings, with high earnings in one year followed by much lower earnings in the next.

Victoria University of Wellington’s John Creedy and Norm Gemmell delved into tax data to see what happens when we follow individuals over time rather than just tracking how much money is earned by the person at the top of each decile band. In the charts below, the red bars follow the same method used in the traditional ‘snapshot’ cross-sectional analyses. Everyone is lined up from lowest income to highest income, and the income of the person at the fifth percentile, tenth percentile, and so on, is measured. A different person will be at each position in each year’s snapshot. And growth in incomes at each of those positions is reasonably comparable as we look across the graph.

The blue bars track what happens instead if we follow each person over time. Those who earned the least at the start of the period had the highest subsequent growth in earnings. Those who earned the most at the start of the period saw declines in earnings. It is a very different picture – and especially different from the received wisdom about the rich getting richer and the poor getting poorer.

The combination of mean reversion and the life-cycle of earnings helps explain the effect. Anyone reporting very high income in any particular year is more likely than usual to have had a very good year – and may expect things to ease off in the subsequent year. Those reporting very low income in any particular year are more likely than usual to have had a very poor year – and can hope for better in the year to come.

Those on lower earnings will also be more likely to be younger workers expecting salary increases; those on the higher earnings are more likely to be at the stage in their careers where earnings level off before retirement.

All of this means that income-inequality measures based on annual income snapshots overstate real differences in incomes, depending on what we think matters for equity.

Imagine a world in which each person earned the same amount, in real terms, at each age – but people were different ages. So every twenty-five year old earned the same amount, in real terms, as current fifty year olds earned when they were twenty-five, and today’s twenty-five year olds would go on to earn as much at age fifty as current fifty year olds. Snapshot cross-sectional measures could show sharp inequalities despite everyone having equal lifetime earnings. We are hardly in that world, but the life-cycle path of earnings affects measured income inequality – as do temporary fluctuations in earnings.

Further work by Creedy, Gemmell and Laws shows measured income inequality drops by about a sixth when each person’s income is measured as average earnings over longer periods. Temporary spikes in incomes that too strongly affect snapshot measures of inequality are smoothed out. That finding echoes work showing that inequality in measured consumption is much lower than inequality in measured incomes.

A few bottom lines then, if we wish to avoid knowing things that are not so.

First, measured income inequality has been flat for fifteen to twenty years, depending on the inequality measure used.

Second, while growth in incomes appears broadly shared across the income distribution, annual cross-sectional snapshots looking at income growth at the different deciles miss important parts of income mobility. Tax data from Creedy and Gemmell shows that income mobility matters, and that income growth is strongest among those on lower incomes.

None of this is to say that there are not important problems in inequality in New Zealand. Broken policy around housing markets has created horrible problems where those who owned homes prior to the large run-up in house prices have benefitted, and rents have taken too large a chunk of everyone else’s income.

But it is important to diagnose problems properly – figuring out what is really so – before attempting policy solutions.