Did "all the income gains go to the top"? Below is a tool to test that claim: compare real individual income growth across percentiles for the United States for any time period from 1962-2025. See who actually benefited from economic growth (and when) using the Census Bureau's annual ASEC.
Income growth by percentile calculator
Using the income growth comparison tool
The tool defaults to comparing two 20-year periods. Pick any start and end years for each period, then hit Compare Periods to see growth at every percentile.
Basic comparison
- Period 1 and Period 2: Select start and end years for two time periods to compare side-by-side
- Chart: See annualized real income growth for each percentile in both periods
- Summary: Highlights median growth and top percentile growth for quick comparison
Options and presets
Click Options to access additional features:
- Presets: Quick access to notable comparisons – last 20 vs prior 20 years, a Leonhardt-esque comparison (1962–1980 vs 1980–2014, the closest we can get with CPS ASEC data), first vs second half of the dataset, and pre/post-2000
- Single Period Mode: Analyze just one time period instead of comparing two
- Show All Percentiles: Display every percentile (1-99) instead of the default selection. (Note that aggressive topcoding before 1996 means we won't show the top few percentiles for older dates.)
Reading the results
The chart shows annualized real growth. That's the compound annual rate of income growth, adjusted for inflation. A flat line would mean all income levels grew at the same rate. A line sloping up to the right means higher earners gained more.
Click Detailed Table to see both total and annualized growth for key percentiles in a sortable format.
The debate behind the tool
In 2017, David Leonhardt's New York Times piece "Our Broken Economy, in One Simple Chart" went viral. It compared two 34-year periods – 1980-2014 vs 1946-1980 – and showed a dramatic shift: in the earlier period, income gains were broadly shared; in the later period, nearly all the growth went to the top income percentiles.
The chart sparked a fierce debate. Economists like Auten and Splinter challenged the underlying data, arguing inequality barely budged. Leonhardt's chart was based on work by Thomas Piketty, Emmanuel Saez, and Gabriel Zucman (later, PSZ). PSZ – and supporters – have since called such critiques "inequality denial."
Critics have also pointed out 1980 (a recession year) might be cherry-picked. AEI's Scott Winship has argued that median pay was artificially elevated through the 1970s, so using 1979-1980 as a baseline overstates subsequent stagnation.
(And the debate goes on, feel free to read more and pick a side: one, two, three)
Why our chart looks different
If you run the Leonhardt-esque comparison (1980-2014), you'll notice DQYDJ's chart doesn't show the same smooth upward slope. That's because I use different underlying data:
- Leonhardt used different tax data (Piketty-Saez-Zucman distributional national accounts). We use CPS ASEC survey data - the standard for income statistics, but measuring different things.
- Capital gains are the big difference: The hockey stick in Leonhardt's chart comes largely from the 99th+ percentiles, driven by capital gains. CPS ASEC's INCTOT variable doesn't include capital gains – it covers wages, business income, dividends, interest, and transfers, but not gains from selling stocks or property. For top earners, this can be huge.
- Topcoding: We're capped at the 95th percentile for pre-1996 data due to some aggressive topcoding in our dataset.
- Transfer income: CPS measures Social Security, disability, and other transfers that grew significantly for lower percentiles. PSZ imputes some of these using CPS data, but our treatment differs.
- Income isn't net worth: some authors will try to muddy the differences between the two - I track net worth by year as well, and the Federal Reserve tracks the proportion of household wealth controlled by the top 1%. The proportion of household wealth controlled by the top 1% has undeniably grown quite a bit since the 1990s.
Neither source is wrong, but, obviously, we are measuring slightly different things. The CPS data we're using tells the story of recurring money income; tax-based data captures a picture which includes one-time investment profits.
There's also a conceptual issue with capital gains: they're lumpy.
Sell a house you've owned for 15 years and you might land in the "top 1%" income for that single year – then drop back to your normal percentile (you can't sell it again the next year, at least legally!). Using tax data means you count that as inequality while with the smoother income measure from the ASEC you wouldn't. And which is "right" depends on what question you're asking.
Finally, remember this data isn't longitudinal – we're not tracking the same people in the same percentile over time. The "50th percentile person" in 1980 isn't the same individual as the "50th percentile person" in 2014. People move up and down the income distribution throughout their lives.
What the data shows
I'm not going to tell you what to see – that's why I built a tool, instead. Run some comparisons and draw your own conclusions. The story you find may differ from the conventional narrative, especially since our data excludes the capital gains which drove much of Leonhardt's hockey stick graph.
What I will say: the patterns shift dramatically depending on which years you pick. That's sort of the point; try a few presets, try your own combinations, and see how sensitive the conclusions are to period selection.
Methodology and data notes
The calculator uses individual income data from the Current Population Survey Annual Social and Economic Supplement (CPS ASEC), accessed via IPUMS CPS.
- Income measure: Total personal income before taxes, including wages, self-employment, investments, and transfers. I use the INCTOT variable here, just like with many of my other income posts. Inflation adjustments are from the annual CPI-U snapshots published on this page.
- Topcoding: Pre-1996 data is capped at the 95th percentile due to aggressive topcoding practices. Post-1996 data extends to the 99th percentile.
- Population: Individuals age 16+ with income (same as our income percentile calculator)
Again, note the data is not longitudinal: CPS is cross-sectional – we're comparing "the 50th percentile person in 1980" to "the 50th percentile person in 2014," not tracking the same people over time
Related income tools
Explore more income statistics and calculators:
- Individual Income by Year - Average, median, and percentiles for the whole dataset
- Income Percentile Calculator - See where you rank in the current year
- Individual Income Percentiles - Full breakdown of the most recent year's income distribution
- Household Income by Year - Household income percentiles by year
- Household Income Growth by Percentile - This analysis, household style
- Income by Age - See how income varies across age groups
The "gains went to the top" narrative has become conventional wisdom, but the data is more nuanced than the soundbite. Play with the periods, draw your own conclusions, join the debate, and have fun spinning the dials!
