One Number Will Not Do It

The Gini coefficient runs from zero, where everyone has the same income, to one, where a single person has all of it. It is the number that appears when anybody says a country's inequality rose or fell.

It has a specific weakness, and knowing it changes how you read almost every claim made with it.

The Gini is most sensitive to changes around the middle of a distribution and least sensitive to the tails. So a country where the top 1% doubles its share while everything else stays put can show an almost unchanged Gini. This is not a hypothetical: it describes several rich countries over recent decades, which is why a headline saying inequality was flat and a headline saying the top share doubled can both be accurate descriptions of the same twenty years.

Which one gets printed is frequently a matter of which statistic the writer reached for first.

This is why the serious literature now reports top income shares alongside the Gini (the share going to the top 10%, 1% and 0.1%) because those answer a different question and a summary of dispersion cannot substitute for them.

The second thing that gets conflated is income and wealth, and the gap between them is enormous.

Income is a flow: what you receive in a year. Wealth is a stock: everything accumulated over a lifetime, plus returns on it, plus whatever was inherited. A bad year leaves you with a low income; a bad decade leaves you with debt.

A typical rich country has an income Gini somewhere around 0.3 to 0.4 and a wealth Gini around 0.7 to 0.8. A substantial share of households hold roughly zero net wealth or less, which cannot happen with income. When people quote these interchangeably in an argument, the comparison is meaningless, and it happens constantly.

Thomas Piketty's contribution was to bring long-run data to bear on why the stock behaves as it does, and his shorthand is famous enough to have become a slogan: r greater than g.

If the return on capital exceeds the growth rate of the economy, existing wealth compounds faster than wages can grow, so the capital share rises and wealth concentrates. The historical evidence he assembled, tax records going back two centuries in several countries, shows exactly this pattern before 1914 and after about 1980, with a striking compression in between.

The compression is the part worth dwelling on. Wealth inequality in western countries fell dramatically between 1914 and about 1970, and it fell because of two world wars that destroyed capital, a depression that destroyed more, and then confiscatory top tax rates and strong unions. Piketty's argument is that this period, which many people take as capitalism's normal state, was produced by catastrophe and policy and is not a baseline.

The theoretical claim that r exceeds g by default is contested, with critics arguing that returns should fall as capital accumulates. The empirical claim about the twentieth century is not seriously disputed.

Now the global picture, which cuts across all of this and is the reason arguments about inequality so often talk past each other.

Branko Milanovic's elephant curve plots income growth by position in the global distribution over roughly 1988 to 2008. The shape is a rising back (large gains for the middle of the world distribution, mostly in Asia) then a dip approaching the top, where the lower-middle of rich countries stagnated, then a sharp rise at the very end for the global top 1%.

Both halves are true. Hundreds of millions of people left extreme poverty, which is the largest improvement in material conditions in human history. And the median worker in several rich countries saw almost nothing over two decades. Inequality between countries fell while inequality within many countries rose.

Choosing which of these to cite is usually a choice about which argument you are having rather than a disagreement about facts.

Finally, a measurement problem that makes everything above harder than it looks.

Household surveys miss the very rich almost entirely. There are few of them, they are hard to reach, and they decline to answer. So survey-based inequality statistics systematically understate the top, which is why Piketty and Saez's use of tax records was such a methodological advance.

But tax data captures what is declared. Capital gains are realised when the owner chooses. Income can be retained inside companies. And Gabriel Zucman's work on offshore holdings suggests a meaningful fraction of the world's top wealth sits in jurisdictions that appear in no national dataset at all, which means the measured numbers are a floor, and nobody knows by how much.