This article appears in the Summer 2017 issue of The American Prospect magazine. Subscribe here.Since November, progressives have engaged in many solid post-election audits seeking to explain how the Obama coalition was narrowly supplanted by Donald Trump’s ethno-nationalist wave of white working-class support. Analysts have explored the relative balance of economic versus racial factors; the breakdown in polling, field organizing, and message development; and the role of campaign-specific mistakes and external interventions by the FBI, WikiLeaks, and the Russians. In an election decided by small margins in a handful of states, each of these factors arguably played a role.
Unfortunately, this process has led to some unfruitful debates about whether progressives should double down on the Obama coalition voters, reach out more to white working-class voters, or appeal more to independent and conservative-leaning suburban whites.
When a party narrowly loses the Electoral College it successfully won in two consecutive presidential cycles, and is down hundreds of seats at the state and local level, however, it does not have the luxury of fine-tuning which voters it will reach out to over the next few years.
Democrats need to reach out to all types of voters across a large swath of the country.
The idea that the Democrats should or could afford to ignore white working-class voters, particularly at the state and local level, defies basic political math. Although we’ve written extensively about how the nation’s shifting demographics and ideological attitudes helped to fuel the rise of the Obama coalition, none of these trends preclude the need to build cross-racial and cross-class coalitions in more places.
Younger, more diverse, socially liberal, and Democratic-leaning voters are not evenly distributed across the nation, and even with the long-term expected declines in white working-class populations, these voters alone cannot sustain successful Democratic coalitions going forward. Democrats do not need to win FDR-level support among white working-class voters, but they cannot afford to lose them by margins as high as 30 to 40 points in some key states—as they have in recent elections.
White Working-Class Trends
In 1980, the white working class (WWC) composed about 70 percent of the eligible voter population. As a result of the changing racial composition of the country and the rising rates of educational attainment, the next 36 years saw this group decline by 25 points—down to 45 percent of all eligible voters. While the WWC is still the largest race/education group in the country, it ceased to be the majority of eligible voters around 2010.
Neither the 1980 geographic distribution of this population nor the changes since have been homogeneous, however. As you can see in Figure 1, even in 1980 the WWC population was more concentrated in the nation’s interior as well as the Pacific Northwest and deep Northeast. At the time, only two states had an eligible population that was less than 50 percent WWC—New Mexico and Hawaii.
Fast-forwarding to 2016, the 25 percentage-point national decline of WWC voters as a percentage of eligible voters had an impact just about everywhere. However, the change was especially concentrated along the coasts and in those states with economically attractive metropolitan areas, drawing both newer immigrants and college-educated whites.
Using a more detailed geographical breakdown, Figure 2 displays the percent of the voting-age population that is WWC in each county in the United States. Going deeper than the state level, we can see that the highest-concentration counties are clustered in Appalachia, extending northeast (but shying away from the coast) and northwest through the Midwest, Great Lake region, and the Dakotas.
Although educational projections are difficult, our own calculations suggest that we’ll see declines in the WWC population continue well into the future. Even if we assume that college attainment rates flatten out rather than grow (an assumption that runs counter to current trends), the continuing racial diversification of the nation will cause this group to shrink. We anticipate that the WWC will constitute just 43 percent of eligible voters by 2020, 35 percent by 2040, and 28 percent by 2060.
Figure 1: White Working Class as Percentage of Eligible Voters
Despite the demographic decline that has already occurred for this group and the similarly sized decline likely to come over the next 44 years, the WWC maintains a large political presence in America. Why?
Although a variety of explanations could be proffered, one of the simplest and most compelling is this: The WWC is very well distributed geographically for the purposes of political influence. They are disproportionately concentrated in swing states, as the maps above suggest. But they are also disproportionately concentrated in swing congressional districts. And they are especially concentrated in swing congressional districts within swing states. In Rust Belt states, for example, the typical swing district was 11 points more WWC than the average district across the nation, and 16 points less minority. In short, they live where it counts.
How did these trends affect the 2016 election? We can say a few things with a high amount of certainty.
Whites—including those without college degrees—turned out at a higher rate than they did in 2012. Turnout in the nation as whole was up—probably to the tune of about 1.6 points. Two pieces of data point to rising white turnout as the cause—particularly among WWC eligible voters.
Figure 2: White Working Class as Percentage of Voting-Age Population
First, data recently released in the November supplement of the Current Population Survey—commonly regarded as the gold standard for estimating turnout among specific demographic groups—indicate that white turnout was up overall compared with 2012. Looking at Table 1, we can see that this increase in participation was concentrated among non-college whites.
There is also evidence of rising WWC turnout from election data we have at the county level. Looking at Figure 3, we can see there is a positive relationship between the percent of a county that is WWC and the change in turnout between 2016 and 2012. That is, the counties where the voting-age population was largely WWC are also places that saw increases in their turnout, a pattern that suggests WWC turnout was also up this election cycle.
Clinton made gains among white, college-educated voters while Trump gained among white, non-college voters. Multiple sources of data and accounts confirm that, compared with Romney, Trump improved his margins within the WWC while losing ground with white, college-educated voters. According to the National Election Pool exit polls, these figures were just about a positive 14-point shift and negative 10-point shift, respectively. Assuming those national figures are true, how much of the county-by-county shift in the vote can be accounted for by combining information about the percentage of the voting-age population that is college and non-college white in a given county with the national-level margin shifts? Put another way, if we assumed that all white college and non-college voters everywhere shifted by the amount that was recorded in the national exit polls, how well would that line up with the actual change in vote shares we saw on Election Day? As seen in Figure 4, there is a very strong relationship between this simulated shift and the margin shift we actually saw at the county level.
Turnout Rates in 2012 and 2016 Presidential Elections
Beyond confirming that these data points from the exit polls seem consistent with the actual results we saw on Election Day, this exercise is also useful because it allows us to see where the model fails. That is, while we see a strong, linear relationship between our simulated shift and the actual shift observed on Election Day, it is not a perfect relationship—the data points in Figure 4 fall both above and below the blue line. The question is, are there geographic regions where this simulated shift systematically overestimates or underestimates the actual shift? That is, are there areas where we saw a smaller or larger shift to Trump than we would expect given that national margin shift among the WWC and college whites?
Looking at Figure 5, which displays the residuals from the model displayed in Figure 5, we can see that there are strong regional patterns. The areas in red are places where our model underestimated the shift (that is, Trump did better than we would expect given both the national shift among WWC and college whites and the proportions of these two groups in the county), while the areas in blue overestimated the shift (that is, Trump did worse than we would expect).
Figure 3: Change in Turnout by White Working-Class Percentage in County
As can be seen, there is a significant underestimation of Trump’s gains relative to Romney in the Northeast, the Midwest, the Great Lakes region, and the Dakotas. The relatively low levels of nonwhites in most of these counties, particularly as you get away from the coastal areas in the Northeast, means it is less likely that a change in these populations’ behaviors (turnout or vote choice) could account for this underestimation. This implies that Trump in these areas either a) had larger shifts in his direction among whites, probably due to relatively large margin shifts among white non-college voters, b) benefited from higher relative turnout from WWC populations, or c) some combination of both.
In contrast, the South was generally a region where we saw smaller actual margin shifts than we would have expected given the data from the exit polls. There are a variety of ways to make sense of that data, but the most straightforward explanation is probably the correct one: This was a region of the country where whites were already conservative enough that they couldn’t realistically shift their vote margins further rightward to the degree we observed at the national levels.
Figure 4: Estimated Shift versus Actual Shift
The Southwest, California, and Texas are home to some of the most Hispanic counties in the country. Where we see strong blue patterning, this could suggest a bit of a counterpunch by Hispanics, with potentially higher Hispanic turnout and vote margins offsetting the expected gains we thought we would see given the vote-choice shift among whites. Additionally, we might be seeing something similar to the South—a white population that is already so conservative (though not in coastal California) that it didn’t change as much as whites nationwide.
That said, there is also a strong red streak of underestimation coming up from the southern tip of Texas, hugging the border westward, and extending up into New Mexico. Puzzlingly, we can see in Figure 6 that these are actually some of the most heavily Hispanic counties in the nation.
Figure 5: Model Residuals of Trump Support
How do we make sense of this given the relationships we discussed in the previous paragraphs? Interestingly, the same rural-urban divide that we’ve long recognized as an important cleavage in the white population might have played a role here. Specifically, while Trump likely did poorly among Hispanics as a whole, it’s possible that his margins were somewhat better among this group’s rural populations, a finding supported by analysis of precinct data in Texas. Those highly Hispanic counties that we also identified as having smaller shifts than we’d expect generally have relatively low population density.
As we head into the Pacific Northwest we notice a strong blue pattern of over-estimation in Washington and Oregon despite large white populations. This is in stark contrast to its counterparts in the white and rural counties of the Northeast. It’s possible that college and non-college whites in this notably progressive region of the country shifted, respectively, further away and not as hard toward Trump. Additionally, as we can see in Figure 7, this was also a region of the country where a significant portion of the vote went toward third-party candidates. To the extent that Clinton lost ground compared with Obama in 2012, it would appear to be the case that many voters shifted toward Gary Johnson and Jill Stein.
Figure 6: Percent Hispanic by County
Lastly, Utah is undoubtedly the political outlier of the 2016 election. During the Republican primary, Trump was denounced by Mitt Romney and faced a conservative and Mormon independent candidate, Evan McMullin, who captured an impressive 21.3 percent of the vote. Trump underperformed spectacularly in this state and, as a result, our simple model overestimates the gains he should have made given the demographic composition of the state.
Strategy Going Forward
What do these preliminary findings tell us about Democratic Party strategy and coalition-building in the Trump era?
For starters, rather than debating whether Democrats should appeal to white working-class voters or voters of color—both necessary components of a successful electoral coalition, particularly at the state and local level—a more important question emerges: Why are Democrats losing support and seeing declining turnout from working-class voters of all races in many places?
Figure 7: Third-Party Vote by County
This is just a hypothesis, but in an era of widespread political cynicism, economic and cultural anxiety, and distrust of both business and government, the Democrats allowed themselves to become the party of the status quo—a status quo perceived to be elitist, exclusionary, and disconnected from the entire range of working-class concerns, but particularly from those voters in white working-class areas. Rightly or wrongly, Hillary Clinton’s campaign exemplified a professional-class status quo that failed to rally enough working-class voters of color and failed to blunt the drift of white working-class voters to Republicans.
The Democrats’ strength on social and cultural issues helped them build a national popular-vote majority with high levels of support in deep-blue cities and states. But after eight years of Obama, Democrats were simply unable to make a credible case to working-class voters of all races, in the states and regions that mattered most, that they could deliver on working-class voters’ core economic needs or represent their values and concerns.
Donald Trump played this card perfectly, first taking over his own sclerotic party and then successfully stitching together a targeted Electoral College victory by promising serious change and embodying a completely different approach to politics than either traditional Democrats or Republicans could offer.
Examining the Center for American Progress’s post-election survey, a full 50 percent of Trump voters said the most important influence on their vote in 2016 was that they wanted “to vote for Trump and the chance to shake up the political establishment,” compared with 29 percent who voted mainly for the policy agenda of the Republicans and another 21 percent who said they voted mainly against Clinton. A similar percentage of Trump voters (50 percent) said they strongly agreed with the idea that “Ordinary people’s opinions are more honest and correct than those of experts in politics and the media,” compared with only 29 percent of Clinton voters. Although it’s likely these attitudes did not dominate other racial or economic considerations that drove people to Trump, the overall conditions for populist voting were clear throughout the entire primary and general election cycles.
Democrats can chew over tactical improvements to their campaigns and outreach, but in the absence of a broad party unifier like Obama, they desperately need to reexamine the public face, leadership, agenda, and ideological approach of the party, given this larger populist context. Voters are in no mood for traditional politics carried out by people they feel are out of touch with their everyday needs and values.
The party needs to rediscover its roots as a working-class party, one that was initially exclusionary of people of color but that today can and must represent the interests and values of working people of all races. As the party fights Trump and his brand of divisive right-wing populism, the party needs to bring in more working-class candidates and leaders who can credibly talk with their communities about common economic and social challenges, can forcefully take on the corporate interests that harm these communities, and who can be trusted to fight for the well-being and security of all working men and women.
If not, the Democrats risk ceding the mantle of political change, and possibly losing more elections, to a demagogic billionaire who talks about populist disruption while doing little to help workers and their families, and much to aid his wealthy cohorts.
Click here to read the rest of our series on the White Working Class and the Democrats.