Race · Public Opinion
A Black president and a killing in Minneapolis were supposed to bend the arc of American racial opinion. The survey record says one of them barely registered — and the other moved the country sharply, briefly, and only on one side of the aisle.
When Barack Obama won in 2008, commentators reached for the phrase “post-racial.” When George Floyd was murdered under a police officer’s knee in May 2020, a different prediction took hold: that the country had reached a racial reckoning, a permanent shift in how white Americans understood inequality. Both claims are testable. The General Social Survey and the American National Election Studies have asked Americans the same questions about race for half a century, and they ran straight through both moments. Weight the answers properly and a clear verdict emerges — and it is almost the opposite of what was promised.
Start with the question that gets closest to the heart of it. Since 1977 the GSS has asked whether the worse jobs, income, and housing of Black Americans are “mainly due to discrimination.” To say yes is to locate inequality in the structure of society rather than in individuals. For three decades the share who said yes drifted in a narrow band around a third. Obama’s election did not move it. Then, between the 2018 and 2021 surveys — straddling the summer of Floyd — it jumped 9.4 points, from 46% to 56%, the largest two-survey move in the question’s history.
“Racial inequality is mainly due to discrimination” — by party
The 2018→2021 jump was almost entirely on the left: Democrats +18.1 points, Independents +2.4, Republicans -3.3. Floyd did not build consensus — it widened the gap. And by 2024 the national share had fallen back to 47%, nearly erasing the spike.
This is the shape of the whole story: a sharp, real, but asymmetric and temporary response to 2020. The people already most sympathetic moved the most; the people least sympathetic did not move at all. A reckoning that runs in only one direction is not a reckoning in the country — it is a deepening of the divide already there.
Who moved is clearer still when the same question is split by the respondent’s own race. Black Americans, already far likelier to name discrimination, jumped 15.5 points after Floyd; white Americans moved 8.3. Both rose — but they rose in parallel, leaving the racial gap in perception as wide as it began.
The same question, by respondent’s race
If one fault line was sharpened by Floyd more than any other, it was over the police. The ANES feeling thermometer asks Americans to rate the police from 0 (cold) to 100 (warm). For decades it was one of the least partisan items in the survey — everyone liked the police. Then 2020 split it open.
Warmth toward the police (0–100), by party
From 2016 to 2020, Democratic warmth toward the police fell about ten points while Republican warmth did not move at all. The single most American consensus institution became a partisan object in one election cycle.
Which leaves the other shock — the one that was supposed to matter most. Did electing a Black president warm the country toward Black Americans? The thermometer says no. Warmth toward Black Americans did not rise after 2008; across every partisan group it drifted down through Obama’s first term. Political scientists call it racialization: once a Black man occupied the White House, attitudes toward race and attitudes toward party fused, and the fusion pulled warmth down fastest among the president’s opponents.
Warmth toward Black Americans (0–100), by party
Put the two shocks side by side and they form a single picture of how American racial opinion actually changes. It does not march steadily toward enlightenment, and it does not lurch permanently after a galvanizing event. It moves in the people already inclined to move, in the direction they already leaned, for as long as the event stays in the headlines — and then it drifts back. Obama did not make the country post-racial; he made race partisan. Floyd did not make the country reckon; he made the half that was already listening listen harder, for a while. The mirror each held up reflected the divide that was there before.
Sources: General Social Survey (1972–2024) and the ANES Cumulative Data File (1948–2024). All estimates are survey-weighted (GSS person post-stratification weight; ANES cumulative weight). Shares and thermometer means are computed within year×group cells of at least 60 respondents; bands in the underlying data are bootstrap 95% intervals. “Party” folds leaners into Democrats and Republicans. The GSS conducted no 2020 round; its post-Floyd reading is the 2021 survey.