The American Mirror · two views, one society
MirrorUS
Shrink any group to 100 people and stand it next to another. Does Congress look like the country? Do the people on screen look like the people watching?
▶ Open the mirrorThis build runs on fictional placeholder numbers, shaped like the real sources they stand in for (Census ACS, BLS, BJS, CRS, league reports and others). Use it to learn how the comparison works, not to quote figures. Verified datasets will replace the placeholders one by one, each with its source attached.
Two rooms of 100
Huge numbers are hard to picture. "If the U.S. were 100 people" is easy. The mirror draws two groups that way, View A and View B, from 66 categories in 11 areas: population, government, business, professions, justice, pro, college and high school sports, and film, music and news.
Between the two rooms sits the spine: for each group, B minus A in percentage points. If a group is 13% of the country and 22% of a profession, the spine reads +9. That is 9 points, not 9 percent: the group's share there is about 70% larger than its share of the country.
Five groups that add to exactly 100
To compare two rooms, both have to be cut the same way, and everyone has to land in exactly one group. The mirror follows the Census convention: Hispanic origin is asked separately from race, so Hispanic means of any race, and the other four groups are non-Hispanic.
The colors mark categories, not skin tones. People choose their own categories in nearly every source here, and those choices can change between surveys.
Why 33 + 33 + 33 doesn't make 100
Nobody can draw 33.3 people. Rounding each share on its own often leaves you with 99 or 101 figures. The mirror uses largest-remainder rounding: every group first gets its whole part, then the leftover seats go to the groups with the biggest fractions until the room holds exactly 100. Type some counts and watch.
| Group | Exact | Rounded | Largest rem. |
|---|
At 1,000 figures each one is a tenth as big, so the rounding error shrinks too. The info button on each mirror lists every group that rounding moved.
When the mirror won't draw
A picture of 100 people looks precise whatever went into it. So when the data can't support the picture, the mirror shows a notice instead.
Rates aren't shares
An imprisonment rate is people per 100,000 of each group. Those numbers don't add up to anything, so the mirror shows them as bars and leaves the spine blank.
Categories that don't line up
Some sources report race but not Hispanic origin. Squeezing them onto the five-group scale would invent a comparison, so they're shown as not comparable.
No guessing
If a source has no women-only figures, the mirror won't estimate them from the totals. Switch to All to see what the source reports.
Time, not heads
On-screen speaking time is a share of minutes. There, one figure means 1% of the time, not one person.
Which room do you compare against? Comparing a league to the U.S. population is a choice. Many baseball and hockey players were born abroad, so the U.S. population isn't a like-for-like baseline for them. Children, workers, students and voters are all different baselines too. Pick View A with care.
Controls
- Views
- Tap a mirror's title, or A or B in the dock, to choose its category. Search finds any of the 66. The arrows swap the two.
- Display
- Split by sex (women fill left from the center line, men right), dots, figures, or figures shaped by sex. Sex shapes appear only where the source reports sex.
- Sex
- All, M or F, for sources that break it down.
- Highlight
- Tap a color in the legend or a row on the spine to light up one group in both rooms.
- Info ⓘ
- Source, year, denominator, exact shares, rounding, method and data quality for each view.
- Settings
- 100 or 1,000 figures, and letter labels on figures.
Best in landscape on a phone.
Demo data. Group definitions follow the U.S. Census Bureau's race and Hispanic-origin questions.