Once Ohio, North Carolina and Florida voted on March 15, the Democratic nominating contest was essentially over. In the minds of some, it's currently not over, even at this point, but that's mostly some mix of hope, rhetoric and denial.
By March 15, it was clear 1) that Hillary Clinton had run up a big enough lead that Bernie Sanders couldn't catch up with her without a big shift in how Democrats were voting, and 2) that the pattern of voting in the Democratic contest hadn't changed. Young voters were voting for Sanders, old voters for Clinton; Democrats for Clinton, independents for Sanders.
And then there were black voters. Black voters were supportive of Hillary Clinton by a wider margin in 2016 than they were of Barack Obama over Clinton eight years prior. A big factor in Clinton's wide delegate lead by March 16 was that she'd swept the South by huge margins, thanks to heavy black turnout and wide margins of victory.
We were curious whether or not black votes were the key to Clinton's victory -- that is, if Bernie Sanders would have won if the black vote hadn't been as strongly supportive of Clinton. That's a hard question to answer, but here goes.
We have data from exit polling in most of the states that held Democratic contests which tells us how much of the electorate was black and how those voters voted. There are big gaps in the data, which we'll get to, but it gives us a decent sense of how changes in the black electorate might have changed the race.
By our rough estimates, each shift of the black vote toward Bernie Sanders by 5 percentage points in each state would have narrowed Clinton's lead by about 31 delegates. Since there are no exit polls from California, and since the state had the most delegates of any on the calendar, we estimated the composition of the vote there using a poll from the Los Angeles Times and USC that was released earlier this month. It ended up being the poll closest to the actual results.
Democratic delegates are allocated proportionally by state and congressional district. Without breaking down the vote in each congressional district, we simply applied the new percentages after shifting the black vote to the statewide totals, which ended up giving Clinton slightly more delegates (suggesting that the by-district system helps Sanders). So remember that all of the margins in our estimates are probably a bit higher than they would actually be.
If the black vote shifted 50 points toward Sanders -- i.e. if Clinton won the black vote in a given state by 55 points, her adjusted margin would now be 5 points -- by our estimates Clinton would still have an overall delegate lead of about 65 delegates. If they split the black vote evenly in every state, her lead would still be more than 30 delegates -- a much, much smaller lead, but still a lead.
That's shifting only states for which we have exit polls (and California), but many of the states for which no exit polls are available are also heavily white, rural states.
How is that possible? Simple: Over most of the course of the campaign, Clinton led or was tied with Sanders among white voters, according to Post/ABC News polling. The important exception was in our March poll, the point at which the two were also the closest in national polling.
In the average of exit poll results among white voters -- which doesn't include every state -- Sanders has earned 49 percent of the white vote to Clinton's 48.
But of course there are a lot of Democrats who are not white or black. Hispanic* Democrats also preferred Clinton, and that is why she would almost certainly have won anyway, even if black voters had gone 50-50. In exit polling, non-white voters (including black and Hispanic voters) went for Clinton by a more than 2-to-1 margin. In December, Gallup found that Asian-American Democrats viewed Sanders more favorably than Clinton; that Times/USC poll had Sanders leading with that group. But Hispanics make up much more of the Democratic electorate than do Asian Americans.
It's fair to wonder whether or not a much-closer race among black Democrats and, therefore, a much-closer delegate fight over the last two months might have shifted other dynamics in the race. This is all speculation and estimates, and how a change of that nature would have reshaped what might have happened is also speculative. What this exercise is perhaps the best reminder of is that Sanders didn't make much headway with black voters, and ended up losing relatively handily as a result.
* Here, we're using exit poll differentiations on race, which make "white," "black" and "Hispanic" exclusive in a way most demographic analysis doesn't.