If you squint you can see it.

Photographer: AAMIR QURESHI/AFP/getty images

The World's Most Important Graph Heads to the Graveyard

Noah Smith is a Bloomberg View columnist. He was an assistant professor of finance at Stony Brook University, and he blogs at Noahpinion.
Read More.
a | A

Most human beings suffer from what psychologists call confirmation bias. When we hear a story that we already believe is true, we trust the storyteller even more. In the age of big data, the story often comes in the form of a graph. And so it is that most of us in the economics profession and financial press find ourselves with egg on our faces. The so-called Elephant Graph, which many of us had labeled the most important in the world, probably doesn’t show what it claims to.

The graph, made by development economist Branko Milanovic, shows how the income levels that define the percentiles of worldwide income distribution have changed since globalization kicked into high gear in about 1980:

This appears to show that the gains from globalization went mostly to rich people in developed nations and to poor and middle-income countries like China, India and Africa, while the middle and working classes in developed countries languished. It turns out that while that general story is probably more or less true, this graph doesn’t really show it.
QuickTake Income Inequality

There are several big problems with the elephant graph. Many of these were discovered by the Resolution Foundation, a British think tank, and are explained in wonderful detail in a report by researcher Adam Corlett.

The first problem is that the graph doesn’t adjust for population. The Elephant Graph measures income at various percentiles of the distribution, but that distribution changes a lot due to population growth. If poor countries grow faster than rich ones, it can create the illusion of increasing inequality.

To see this, imagine a world in which there are just two countries: Richlandia, where everyone makes $100,000 a year, and Pooristan, where everyone makes $10,000. Richlandia has slightly more people than Pooristan, so the global median income is $100,000. Now imagine that both countries see their income double, but that Pooristan’s population grows faster, so that now it has slightly more people than Richlandia. Now, even though everyone in every country is much richer than before, global median income has fallen from $100,000 to $20,000 -- a drop of 80 percent! A global incidence curve like the Elephant Graph will show this drop, but it’s an illusion.

The Resolution Foundation’s report shows what happens if you adjust the graph for population growth. They also show how it changes if you exclude just two or three countries from the graph. Lo and behold, the elephant disappears:

The peak of the elephant’s “head” in the Elephant Graph turns out to be mostly due to China -- a point also made by David Rosnick of the Center for Economic and Policy Research. That’s probably not surprising, given that China contains a fifth of humanity and has been a star performer in terms of economic growth.

But even more damning for Milanovic’s elephant is that much of the “trunk” -- the poor performance of the global upper middle class -- turns out to have been due in large part to the collapse of the Soviet empire. The economic meltdown that devastated incomes in Eastern Europe and Russia in the 1990s is surely a human tragedy, but it isn’t related to the stagnation of middle-class incomes in the U.S. or Europe. Most of the rest of the dip in the Elephant Graph turns out to be Japan’s 1990s stagnation -- again, a bad thing for the world, but not the story we were using the graph to tell.

Meanwhile, Francois Bourgruignon of the Paris School of Economics has developed a better method of displaying global growth. This is what he calls “non-anonymous” incidence curves, which try to measure the income growth of people based on where they start off in the global distribution. So far, Bourguignon has only produced these graphs for a few short periods of time, but they look much more flat and less dramatic than Milanovic’s elephant.

So why were we so quick to use the Elephant Graph to explain everything in the world? Probably because we already believed in the broad contours of the story we were telling. We know that since 2000, median income has stagnated in rich countries. For example, here’s the U.S.: 

Worse off
Median household income, adjusted for inflation
Sources: Federal Reserve Bank of St. Louis

We also know that China has grown like gangbusters since about 1980, and India has posted solid gains since the early 1990s. Here is Indian per capita gross domestic product:

The Right Direction
India inflation-adjusted per-capita GDP*
Source, World Bank, Bloomberg
*2005 prices

We also know that poverty has fallen all over the world since 2000, and that inequality between rich and poor countries has also decreased since then. And we know that the richest of the rich -- the 0.1 percent in developed nations -- have reaped enormous gains during that time.

So the story we were using the Elephant Graph to tell is a good story, at least for the last 15 years or so. The problem is that the Elephant Graph doesn’t actually tell it. Instead, it’s a lucky accident of data visualization that confirmed our pre-existing beliefs.

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

To contact the author of this story:
Noah Smith at nsmith150@bloomberg.net

To contact the editor responsible for this story:
James Greiff at jgreiff@bloomberg.net