When the Rich Get Richer and the Poor Get Poorer, That Is a K-Shaped Economy. Are We in One?

The letter K has been turning up in every second economic argument for months. Usually without explanation, often as proof that things are worse than the numbers say. The term is more precise than its use suggests — and in July 2026 the Fed tested it properly for the first time. The result is more interesting than the headline.

The letter comes from an old habit

Economists have described recoveries with letters for decades. A V drops steeply and climbs back just as steeply. A U drops, sits on the floor a while, then returns. An L drops and stays down. A W drops twice.

Each of those letters draws one line. The K draws two.

It surfaced on Twitter in April 2020, from an anonymous account called Ivan the K. Behavioural economist Peter Atwater picked it up and wrote about it through the summer; by August 2020 the term was everywhere. Atwater did not arrive at it through wealth data but through sentiment: people who could move to working from home regained confidence within weeks. People standing in a hospital, at a supermarket till or on a factory floor kept losing it.

There is no agreed definition

This is where the trouble starts. The Richmond Fed says so in the opening lines of its July 2026 brief: no single accepted definition exists. The available analyses differ in how they group households and in what they measure at all.

So the authors commit to one, and a strict one. K-shaped means: outcomes improve for high-income households and worsen for low-income ones. Not more slowly — backwards. An economy where low incomes grow 2 percent and high incomes grow 6 is explicitly not K-shaped under this definition.

Households are ranked by income and split into fifths: bottom 20 percent, top 20 percent, the middle 60 percent in between. The measure is income after taxes and transfers, because that is what a household actually has to spend.

Loosen the definition to something like "the gap is widening" and almost every year since 1994 qualifies. Keep it strict and few do. The argument about the shape is largely an argument about the definition.

What comes out

Real after-tax income across three decades, 1994 to 2024:

All three are up. Under the strict definition the full period is therefore not K-shaped. The gap still widened substantially — the two statements do not conflict.

Two stretches look different. Between 2002 and 2006, and again between 2010 and 2013, real income for the bottom group fell while the top group's rose. Both are recoveries from a recession.

After the COVID recession it did not happen. The three groups moved together, supported by government transfers. Anyone dating the K from the pandemic is dating it from the one recession where income did not split.

Where the money comes from

Breaking income into sources explains why.

For high-income households, earnings are both the largest source and the largest driver of growth. Taxes and transfers net out negative for them — they pay in more than they take out. The fastest-growing line was the remainder: interest and dividends on assets.

For low-income households the picture inverts. Across the period, more than half of their income is transfers minus taxes. And their growth between 2014 and 2024 came almost entirely from that line rather than from wages.

One detail from the same breakdown says a lot about the episodes: between 1999 and 2004, real earnings grew for the top group while shrinking for everyone else.

Income and consumption give different answers

What a household earns is not what it spends. Consumption is far less dispersed than income — households smooth, borrowing or drawing down savings, and the tax and transfer system redistributes.

The authors flag one caveat themselves: the underlying expenditure survey probably undercounts spending by high-income households, concentrated in luxury goods. Correct for that and the dispersion widens considerably.

In consumption, faint K-patterns appear in 2002 to 2004, 2011 to 2013 and 2021 to 2023. The last of those is the notable one: consumption was K-shaped there while income was not.

Newer figures, assembled by the New York Fed from retail data collected by the firm Numerator, run from early 2023 into the first quarter of 2026. They show a K in 2023, then a uniform rise across all income groups from 2024 — and a decline in the most recent period.

Why short stretches last

Here is the finding usually missing when the letter gets used. K-shaped episodes are mostly not followed by steep recoveries. What is lost in one of them largely does not come back.

Then there is mobility. If moving up and down were random, nobody would sit permanently on one arm of the K: over a lifetime a household would spend roughly as long at the top as at the bottom, and its average would land in the middle. In practice, US income mobility is low. Whoever is on an arm is likely to stay there.

Put together, that is the real claim: you do not need a permanent condition to produce permanent differences. A few years are enough if nobody catches up afterwards.

Why an average can never show it

A K cannot be read off any single average. Not GDP, not median income, not the inflation rate. That is not a data-quality problem but a construction one: two groups moving in opposite directions produce an average that barely moves. The split vanishes into the mean because a mean is built to do exactly that.

Which is why distributional data exists — the household survey by fifths, the expenditure survey, the Distributional Financial Accounts. Those datasets were built for this.

The same holds for inflation. One rate hits different baskets differently. The number in the release is an average across spending patterns that differ sharply between groups.

The same mechanism sits inside every index

A stock index is an average too. A market-cap-weighted one is a weighted average: it tells you about the largest members, not the typical one.

An example from this year. In the first half of 2026 the seven large tech names were collectively flat, while the S&P 500 rose 9.3 percent over the same window. In 2023 and 2024 it ran the other way: a narrow group carried nearly all of the move and the rest barely joined.

Two completely different markets. The index level alone cannot tell you which one is running. For that you look at breadth — the same index equal-weighted against cap-weighted, for instance. That is nothing more than asking for the distribution instead of the mean.

A backtest poses the same question again. An average return over five years does not say whether it came from three trades or thirty. It does not say how far the equity curve fell in between, or for how long. Those two figures — drawdown depth and duration — are distributional, and they decide whether a strategy is one you can actually hold. The mean never decides that.

The argument is still open

Whether the US economy is currently K-shaped remains contested. The Bank of America Institute sees the card-spending gap between income groups narrowing since May 2026 and describes convergence rather than a split. The Minneapolis Fed reviewed the consumer data separately the same year. Other letters are in circulation.

None of that is a reason to drop the term. It describes a real pattern and points at the right question. It is only worth asking three things every time it appears: under which definition, from which data, over which window. The letter is a description, not a measurement.

---

Not investment advice, not a forecast.

Sources: Federal Reserve Bank of Richmond, Economic Brief No. 26-23, July 2026, "Is the US Economy K-Shaped? Evidence From the Past Three Decades" (DeMaria, Jones, Mengedoth, Neelakantan) · Federal Reserve Bank of Richmond, Speaking of the Economy, 17 June 2026 · Federal Reserve Bank of New York, Liberty Street Economics, May 2026 · Federal Reserve Bank of Minneapolis, Jeff Horwich, 2026 · Peter Atwater, "The K-Shaped Recovery: A Narrative Economics Case Study", September 2020 · CNBC, 23 and 29 August 2026

Study the Past — Improve your Future 🥋