Even before the election, political analysts were decrying Hillary Clinton as a candidate. She wasn’t a natural campaigner. She wasn’t spending enough time on the ground. She had too much baggage. She couldn’t break through the noise about her email server. The natural extension of this argument is that a typical Democrat would have performed better than Clinton did and will in the future. However, the election results in key battleground states do not support this theory. Democrats lost up and down the ballot. And they have been losing cycle after cycle in non-urban America for some time.
By the middle of the 2016 cycle, there were nine competitive Senate races around the country: Florida, Indiana, Missouri, Nevada, New Hampshire, North Carolina, Ohio, Pennsylvania, and Wisconsin. Heading into 2016, Democrats felt confident about this favorable map, but Republicans ended up winning seven of those contests. In New Hampshire, Democrats staggered to a narrow 1,017-vote (0.14 percent) victory, and only in Nevada did the party record a healthy win. Notably, in both of these two states, Clinton had a clearer ground game advantage over Trump than in the places Democrats lost.
Democrats also came up short in state legislative races around the country. Of the 7,383 state legislators in America, Republicans grew their margin by 90 seats in 2016. Building off of big Democratic losses in 2014 at the state level, Republicans currently occupy 1,038 more state legislative seats than Democrats nationwide. They now control 68 of 99 legislative chambers and 32 of the 50 governorships.
These massive losses up and down the ballot indicate problems that are much bigger than any weakness Clinton might have had as a candidate or any drag she might have had on the rest of the ticket.
Democrats have been losing small-town, exurban, and rural voters cycle after cycle when Barack Obama has not been on the ballot. The consistency of those losses is illustrated in recent election results from the five states that Working America surveyed: Michigan, North Carolina, Ohio, Pennsylvania, and Wisconsin.
Democrats have been losing small-town, exurban, and rural voters cycle after cycle when Barack Obama has not been on the ballot. Here, Obama campaigns in Merrimack, New Hampshire, in 2008.
In the 2014 Michigan gubernatorial race, unsuccessful Democrat Mark Schauer received 43.09 percent of the vote in the non-urban counties. Clinton’s 2016 performance fell a bit further behind his result (38.26 percent). In contrast, Obama received 49.28 percent of the vote in these same counties in 2012. Both Schauer and Clinton lost the state, yet Obama won it twice.
In the 2010 and 2014 Wisconsin gubernatorial races, unsuccessful Democrats Tom Barrett and Mary Burke received 41.61 percent and 41.27 percent, respectively, in non-urban counties. Clinton’s 2016 performance tracks slightly behind these benchmarks (39.38 percent) and was 7.5 points behind Obama (46.77 percent). This difference was a large part of the reason that Obama won Wisconsin twice while other Democrats lost.
This same pattern held true in the North Carolina, Ohio, and Pennsylvania midterm elections. Since 2008, Democrats have resoundingly lost the elections where Obama was not on the ballot, with a substantial share of those losses coming from a lack of support from working-class voters, especially in non-urban areas.
Much post-election analysis has focused on the drop in turnout among African American voters. While we think lower turnout in this bedrock of the progressive base was a factor in some places, African American turnout was not down everywhere. It did dip in crucial cities like Detroit and Cleveland, but turnout was up in other heavily African American communities like Greensboro, North Carolina. Where there was a drop in turnout, it was not decisive for most states and in no state was it the largest share of vote loss.
Some election postmortems hold that if we can just return African American turnout to 2008 or 2012 levels, then Democrats will win again. But higher African American turnout in the battleground states would not by itself have led to a Clinton victory. The votes are simply not there. To test this theory, Civis Analytics looked at levels of support in high-percentage African American counties in our five states as well as in Florida. It categorized high-percentage African American counties as any county with an African American population of 25 percent or more. In each of those identified counties, it replaced Clinton and Trump raw votes with Obama and Romney votes, respectively, and calculated the new state-level results.
Out of the six states, only Michigan (16 electoral votes) and Ohio (18) flipped to Clinton. While these two important states would have made the Electoral College count a lot closer (272-266), the national outcome would have remained the same. This research suggests that while African-American turnout is an important focus, our overall strategy must also include outreach to persuadable white working-class voters. We must reach them in non- urban communities and speak to their unique social and economic anxieties.
In fact, Democrats’ electoral efforts would be substantially improved by more investment in political organizing that engaged both communities of color and white working class voters at all times, not just in the few months before Election Day.
CLINTON WON 1.02 MILLION fewer votes than Obama in the five states we examined. Our analysis of this drop-off found that it was driven largely, but not exclusively, by white Obama voters who swung away from Clinton and who lived disproportionately in non-urban areas. Of the votes lost, 831,082 (81 percent) came from non-urban counties where the population was considerably whiter than the urban centers.
In a separate analysis, Working America examined our candidate ID data from 7,531 voters we canvassed in Ohio, Pennsylvania, and North Carolina in both 2012 and 2016, focusing on the 4,854 voters who supported Obama. (These IDs were concentrated largely in urban population centers and are most indicative of the trends in those communities.) This ID data provides insights into voters who supported Obama in 2012 but ended up backing Trump in 2016.
Of the 3,799 white voters who voted for Obama in 2012 and whom we canvassed in 2016, 28 percent told our canvassers that they were undecided or supporting Trump in 2016. However, white voters weren’t the only defectors. Of the 1,074 African American Obama voters canvassed in 2012 and again in 2016, 12 percent were undecided or supporting Trump in 2016.
We also examined voter file records in these five states to answer the question: Did 2016 produce a surge in angry white voters supporting Trump?
The evidence from these completed 2016 voter file suggests that while there were increases in turnout from white conservatives, there was not a determinative surge from new, disgruntled white voters. Rather, the data show existing voters swung away from Democrats. We examined what share of voters in urban counties (which tended to support Clinton) and non-urban counties (which tended toward Trump) had cast ballots in 2012 or were new to the 2016 cycle. In these data, the share of new voters in urban counties and non-urban counties was nearly identical to 2012 levels, indicating that much of the change in the electorate was part of the normal churn of new voters participating and former voters dropping out. Using county-level election results, we compared the geographic distribution of new voters and found they were not overwhelmingly more Republican-leaning than the existing voters. In North Carolina, for example, the increase in the white share of the electorate was as likely to be in places where Clinton’s vote increased as in places where Trump’s increased compared to 2012. In Ohio, where turnout was down for white voters, albeit by a lesser amount than for black voters, the new voters were as likely to be in the urban Clinton strongholds as they were to be in non-urban Trump communities.
Taken together, these data points indicate that a significant number of voters who backed Obama in 2012 did not back Clinton; that these voters were not exclusively white; and that discontented white voters who were not already part of the anticipated electorate were not a critical factor in Trump’s victory. It supports the idea that many of the voters lost by Democrats are not racially animated Trump enthusiasts, but working-class people available to be reached by progressives.
These findings jibe with a recent survey by Hart Research’s Guy Molyneux, which estimated that across America there are 23 million white working-class moderates who are open to progressive ideas. His research showed that Trump beat Clinton among white working-class conservatives by an overwhelming 85-point margin. The outcome was much different among white working-class moderates, whose support Trump won by 26 percent margin. In 2012, that same group of moderates voted for Mitt Romney over Barack Obama by just 13 percent. Had Hillary Clinton not suffered such a steep drop in support from white working-class moderates, she almost certainly would have won Michigan, Pennsylvania, Ohio, and, with those states, the presidency.
WHETHER BLACK OR WHITE, Latino or Asian, working-class Americans have suffered in the modern economy. While the divergence of wage and productivity growth has continued for more than 40 years, the trends accelerated for working Americans in the decade from 2004 to 2014.
During much of the period from 2004 to 2014, job growth did not keep pace with population growth, and labor-force participation declined. While the number of working-age people increased during that decade, the share who were working or actively seeking work, regardless of race, decreased from 66 percent to 62.9 percent, a drop of 3.1 percentage points. This drop in labor-force participation, the equivalent of 7.7 million fewer people working, was concentrated among people ages 16 to 54, while older workers held onto jobs and remained in the workforce longer.
Though the economy has been harsh to most working-class Americans, the political response of the white working class has been distinct from other demographic groups, and it’s been even more pronounced in non-urban communities. Even as partisan support for Democrats from African American, Asian, and Latino voters was either steady or grew from 2006 to 2016, it has been dropping steadily among white working-class voters in cycle after cycle. And in 2016, they swung dramatically away from their 2012 levels of support for Barack Obama and other Democratic candidates, especially in non-urban communities.
What explains the unique political response of white working-class voters? To understand the extent to which unique economic experiences might play a role, we focused on three economic measures in our battleground states—wage growth, job growth and employment levels for prime working-age adults. We looked at whether there were clear differences between urban counties and whiter non-urban counties. While this imperfectly captures differences between black and white workers, it does get at some of the different lived experiences of people in these areas.
For two of these economic measures, it was hard to find clear and consistent differences between the urban and non-urban counties:
A quick look at wage growth in our five battleground states might make one think that non-urban areas were clearly better off. Wages did increase more rapidly in non-urban counties than urban ones over the last 10 years. But dig a little deeper and one sees that non-urban wages are actually lower than those in urban population centers.
Within the five states, job growth was a percentage point higher in urban areas (2.45 percent) than non-urban areas (1.4 percent) over the last decade. So that might lead one to believe that non-urban voters faced tougher economic prospects. However, the picture gets more complicated the closer you look. In Ohio, urban areas actually saw a decrease in the number of jobs over the last 10 years. And in Michigan, both urban and non-urban areas lost jobs in the decade past.
But there was one economic measure that did show a clear and consistent difference between urban and non-urban areas: the percentage of prime working-age people with jobs.
We found a clear distinction between urban and non-urban areas when we compared the total number of working-age people (those 18 to 64) to the total number of jobs in the five states we analyzed. While there are people both younger and older who participate in the labor market (they’re included in the labor-force participation statistic cited at the top of this section), this measure of prime working-age adults reflects the large majority of the potential workforce. And it is in this comparison that we see a stark divergence in the economic vitality of urban and non-urban population centers.
In 2015 in the urban counties of our five states, there were 12.4 million people employed and 15.9 million people in the 18-to-64 age range, a ratio of 78 jobs for every 100 prime working-age adults. In the non-urban counties in our five states, there were 9.6 million working-age adults employed and 15.1 million 18- to 64-year-olds, a ratio of 64 jobs for every 100 prime working-age adults. The numbers reveal a much starker employment picture in the non-urban counties. To put the difference in context, were the ratio of jobs to prime working age adults as high in non-urban areas, 2.1 million more people in these five states would be employed.
While this comparison has its limitations, the data suggest that in each state the likelihood of prime working-age adults being employed is far lower if they reside in a non-urban community than if they live in a big urban population center. This factor could be one of many economic pressure points that helps explain what is pulling these communities rightward politically. Immediately following the election, new data emerged correlating economic distress, mortality and morbidity rates, and Trump support, adding another important dimension to the striking political difference separating white working-class Americans from other demographic groups.
In recent years, there has been a startling rise in mortality rates among middle-aged white Americans. This trend was initially flagged in Anne Case and Angus Deaton’s pivotal 2015 study showing that, contrary to every other racial, ethnic, and age group, the death rate for middle-aged white Americans has been rising.
What’s notable for the 2016 election analysis is that the white non-urban death rate is rising faster than the white urban death rate. Back in 2004, there was already a stark difference between the death rate for non-urban and urban whites as a percentage of the total population (0.91 percent for urban and 0.98 percent for non-urban). In the 10 years that followed, the urban death rate increased by 0.03 percentage points to 0.94 percent, and the non-urban death rate increased by 0.08 percentage points to 1.06 percent. This non-urban increase is more than double the rate in urban counties.
In a recent research brief, Professor Shannon Monnat from Penn State University explored the relationship of death rates and economic prospects as they related to Trump support. Her research has profound implications for progressives hoping to reach persuadable Trump voters in future elections. She found that Trump “overperformed the most in counties with the highest drug, alcohol, and suicide mortality rates” and “performed best in counties with high economic distress and a large working class,” and that these two factors are related. Indeed, the dramatic rise in death rates, linked largely to high drug, alcohol, and suicide rates, is accounted for by economic distress, namely the decades-long erosion of the job base in these regions. She goes on to note, “What these analyses demonstrate is that community-level well-being played an important role in the 2016 election, particularly in the parts of America far-removed from the world of urban elites, media, and foundations.”
THE SHIFTING COMMUNICATIONS LANDSCAPE undoubtedly abetted the outcomes in this election. But Democrats’ campaign spending choices were not calibrated to address this changing landscape, especially given the challenges of reaching unsettled white working-class voters in battleground states.
While African American turnout dipped in crucial cities like Detroit and Cleveland, but turnout was up in other heavily African American communities like Greensboro, North Carolina. Here, residents of nearby Raleigh line up for early voting in October 2016.
The primary way most voters received information about the election was via cable and broadcast media as well as social media. Trump was the clear beneficiary. Just the difference in total volume of media exposure between Trump and Clinton was larger than all campaign-directed communication (paid ads, direct mail, field) combined. From the outset, Trump’s celebrity drove an outsized advantage in the free media he received from traditional and social media outlets—an advantage over Hillary Clinton of an estimated $2 billion.
Trump’s free media advantage was evident early on, so Clinton’s backers were confronted with a question of how to counteract their disadvantage with paid campaign outreach. Here Clinton had a clear war-chest advantage. In just the last four months of the election cycle, Clinton spent $211 million on TV ads in battleground states, almost triple the $74 million spent by Trump—and twice the advantage Obama enjoyed over Mitt Romney in his successful 2012 campaign.
This pattern of robust paid-media spending was consistent with long-standing practice in progressive politics. According to a 2014 article by researchers David Broockman of Stanford University and Josh Kalla of UC Berkeley, the large majority of campaign spending is centered on TV and digital advertising, with only a fraction going to direct voter contact.
The share of investment that did go to direct voter contact was focused to a great extent in the urban population centers, which mostly targeted progressive base voters in the last few months of the campaign. According to one cumulative tracker for the progressive community, 80 percent of all contacts with voters in the five battleground states took place in the urban areas where Clinton did best; 77 percent of contacts targeted Clinton’s base, and 83 percent occurred after Labor Day.
The result was that most voters in battleground states did not recall any direct voter contact. According to the national exit polls, only 43 percent of Ohio voters, 38 percent of North Carolina voters and 51 percent of Pennsylvania voters reported receiving direct voter contact from the campaigns or their allies.
Looking at the same question by candidate, 28 percent of Ohio voters said they were contacted by the Clinton campaign, while 24 percent reporting contact from Trump’s. In Pennsylvania, 37 percent of voters said they were contacted by the Clinton campaign compared with 28 percent by Trump. In North Carolina, 26 percent reported contact by the Clinton camp; 22 percent by Trump. Despite widespread claims that Trump did not have a ground game, recall of contact from his camp was nearly as large as that from Clinton’s. Notably, this level of direct voter contact was down from the 2012 cycle, when 60 percent of battleground-state voters reported receiving direct voter contact. That year just over 40 percent of voters in the five battleground states we examined reported receiving direct contact from the Obama camp. Just under 40 percent reported contact from Romney.
THE SYSTEMIC LOSSES of the 2016 election reflect a profound failure to establish credibility or connect with voters. We need to start winning the trust of working-class voters through year-round, in-person engagement. In those conversations and subsequent communications efforts, we need to change the narrative so voters’ frustrations are refocused on the appropriate targets instead of on other working-class people who are different from them. By doing so, we can defuse right-wing messages that target “others” and negate demands for racial justice.
We must engage voters with face-to-face conversations that are as much about listening as talking. It’s this kind of organizing that persuades the skeptical and mobilizes the committed. As Andrew Cockburn has noted in Harper’s Magazine, “Of all the ways to get people to come out and vote tested by the academics, one emerged as the absolute gold standard. Talking to them face-to-face, the longer the better, turned out to have a dramatic effect. … [T]he effect is infinitely more cost-effective than any traditional media-heavy approach.”
Understanding how and when to move voters toward a candidate is useful, but the task before us in future elections will be much different. Instead of moving voters’ views of a specific candidate, we must acknowledge the right-wing surround-sound in which voters live and utilize an appropriately scaled response to change the way they interpret the shifting political and economic realities. This objective is much more ambitious than changing a vote for a single election. We have the opportunity—and the imperative—to begin this work now.
Click here to read the rest of our series on the White Working Class and the Democrats.