Skip to main content
What Elizabeth Warren’s Last Election Can Tell Us About 2020

One of the many reasons I think New England is great1 is that its elections are conducted — and its results are reported — at the city/town level, not the county level. That lets psephologists like me study them in greater detail, and one election I’m particularly interested in is the just-concluded 2018 U.S. Senate race in Massachusetts.

For Sen. Elizabeth Warren, the race was little more than a tune-up for the 2020 presidential campaign she is now gearing up for2; she breezed to re-election by 24 points (although it was arguably an underwhelming performance given the state’s blue hue). But we can look at who voted for Warren in 2018 for clues as to who might vote for her in 2020 — both in the primary and, if she gets there, in the general election.

This isn’t as simple as just looking at which areas Warren won — because partisanship predicted the 2018 election results so well, all that would tell us is that she does well in the most Democratic places. Instead, you have to look at how Warren performed relative to other Democrats. Since she is vying to succeed Hillary Clinton as the Democratic presidential nominee, I used Clinton’s 2016 general-election performance as my point of comparison. If Warren did better than Clinton in a given town, it suggests that its voters were more enthusiastic about her than they were about Clinton. If Clinton did better than Warren, it indicates that those voters were relatively cool toward the state’s senior senator.

I calculated the difference between Warren’s and Clinton’s vote shares for all 351 cities and towns in Massachusetts as well as the 255 voting precincts in the city of Boston. Armed with demographic statistics about each of those jurisdictions (plus my knowledge as a Massachusetts native), I spent a few days looking for patterns in the data. Here’s what I found.

1. She’s weak in elite suburbs

One clear trend is that Warren underperformed in extremely wealthy, highly college-educated communities. In 2013, The Washington Post conducted a nationwide analysis that identified the wealthiest and most educated zip codes in the U.S., labeled “Super Zips” (building upon an earlier analysis by political scientist Charles Murray). Modifying the Post’s Super Zip calculations to look only at Massachusetts,3 I identified the state’s 12 upper-crustiest cities and towns. The list is a veritable where’s-where of elite Boston suburbs — and Warren did worse than Clinton in all 12.

Warren underwhelmed in ‘Super Zips’

How Hillary Clinton (in the 2016 presidential race) and Elizabeth Warren (in the 2018 Senate race) performed in the 12 wealthiest and best-educated communities in Massachusetts

Vote Share
City/Town Median Income* % with Bachelor’s Clinton Warren Diff.
Wellesley $177k 82.7% 71% 64% -7%
Weston 197 82.4 66 60 -6
Medfield 154 72.5 59 55 -4
Dover 204 82.7 57 54 -3
Needham 142 74.6 70 68 -3
Sherborn 171 82.9 67 64 -3
Winchester 152 75.5 63 61 -2
Sudbury 171 78.5 69 66 -2
Wayland 167 82.6 72 70 -2
Andover 143 73.7 58 57 -2
Lexington 162 81.6 77 75 -2
Carlisle 171 84.7 69 68 -1
Statewide 74 42.1 60 60 0

*By household

Communities qualified as Super Zips if they scored at least a 95 in our calculation, which averaged each community’s percentile in education with its percentile in income to get a score from 0 to 100.

Sources: 2013-2017 American Community Survey, Massachusetts secretary of the commonwealth

Although I’d like to include some of the more well-to-do sections of Boston in this list, no income or education data is available for individual precincts within Boston. To get at least a rough sense of how those neighborhoods voted, however, we can use past election results — specifically, the precinct’s margin shift from 2012 to 2016 presidential results — to find likely Super Zips within the city. (As has been thoroughly documented here and elsewhere, affluent, well-educated Mitt Romney voters flocked to Clinton in 2016.) Seven of the eight precincts that shifted the most toward Clinton are in Beacon Hill or the Back Bay, two of Boston’s poshest neighborhoods.4 Warren lagged behind Clinton in these precincts by anywhere from 2 points to 9 points, confirming our statewide findings.

Of course, after Clinton won upper-class areas by eye-popping margins in 2016, some reversion to the mean is expected. However, it also makes sense that Warren would be unpopular among wealthy voters, given her career-long crusade against big business. It certainly appears that social and economic elites are not part of Warren’s base.

2. She could win back Obama-Trump voters

So where are the voters who love Warren? Here are the 10 cities and towns where she outran Clinton by the biggest margin:

Places that swung toward Trump still like Warren

The 10 Massachusetts cities and towns where Elizabeth Warren (in the 2018 Senate race) outperformed Hillary Clinton (in the 2016 presidential race) by the highest margin

Vote Share
City/Town Median Income % with Bachelor’s 2012 vs. 2016 Margin Shift Clinton Warren Diff.
Hawley $66k 39.6% R+12 51% 67% +16%
Middlefield 75 26.0 R+30 51 67 +15
Cummington 53 43.2 R+11 60 72 +12
Wendell 43 45.9 R+13 69 81 +12
Sandisfield 69 37.3 R+17 53 64 +11
Peru 69 24.5 R+30 50 60 +11
Otis 70 32.5 R+19 47 57 +10
Leverett 87 64.9 R+3 77 87 +10
Tyringham 86 49.3 D+11 69 78 +9
North Adams 39 24.5 R+20 64 73 +9
Statewide 74 42.1 D+4 60 60 0

*By household

Sources: 2013-2017 American Community Survey, Massachusetts secretary of the commonwealth

They have a lot in common. They are all small towns in Western Massachusetts. With a few exceptions, they have incomes lower than the statewide average. Most of them have fewer college graduates than average as well. And Trump improved upon Romney’s margin in all but one of them.

Just like other white,5 blue-collar areas in the rest of the country, Western Massachusetts broke with longstanding Democratic tradition in the 2016 election. Warren’s ability to match or even exceed President Obama’s 2012 performance in these areas suggests that she might be the right candidate to persuade Obama-Trump voters to once again vote Democratic in the 2020 general election. In addition, a majority of Democrats in most of these towns — some of which are fairly bohemian — voted for Bernie Sanders in the 2016 Democratic primary. That’s a pretty good indication that Warren may also find a primary base among economically struggling communities that could be receptive to her populist message.

3. Young people seem to like her

Young people generally make up a small share of the electorate, so it’s hard to draw any firm conclusions about how they might feel about Warren based on Massachusetts election results. So let’s stick to what we know: Warren exhibited moderate strength in Massachusetts’s 10 youngest municipalities.6

Warren’s youth movement

How Elizabeth Warren (in the 2018 Senate race) performed compared with Hillary Clinton (in the 2016 presidential race) in the 10 youngest communities in Massachusetts

Vote Share
City/Town Median Age Clinton Warren Diff.
Amherst 21.4 83% 89% +6%
Sunderland 31.3 71 76 +5
Aquinnah 25.6 82 86 +4
Somerville 31.5 83 87 +4
Williamstown 29.9 81 84 +3
Cambridge 30.4 88 91 +3
Springfield 32.9 74 75 +1
Boston 32.0 81 81 0
Lawrence 31.4 82 81 -1
Wenham 24.7 56 54 -3
Statewide 39.4 60 60 0

Sources: 2013-2017 American Community Survey, Massachusetts secretary of the commonwealth

In six of them, Warren outran Clinton by 3 points or more, and in three of the other four, more than half the population is nonwhite, which is likely far more important in explaining the communities’ electoral preferences than age is. (The 10th town, Wenham, is an easily-explained-away exception: It’s an elite suburb that appears on the list mostly because a local college7 drops its median age.) Although the sample size is small, this does suggest Warren has a natural constituency in and around college towns, based on her overperformance in places like Amherst8 (+6 points), Cambridge9 (+3 points) and Williamstown10 (+3 points). The Boston precincts where Warren outpaced Clinton the most were also disproportionately located in the city’s Allston and Brighton neighborhoods, which stretch from Boston College to Boston University; as of the last Census, more than half the population there was between the ages of 20 and 34.

4. Nonwhite voters are a wild card

There’s one demographic variable we haven’t mentioned yet, and it’s a big one in a Democratic primary: race. Both the places where Warren did especially well and the places where she did especially poorly were overwhelmingly white.11 How did Warren’s vote share compare with Clinton’s in Massachusetts’s predominantly nonwhite communities (yes, they do exist)? Overall, the differences were minimal, implying that Warren is no better, nor any worse, at wooing these voters than Clinton was.

Warren does OK in diverse places

How Elizabeth Warren (in the 2018 Senate race) performed compared with Hillary Clinton (in the 2016 presidential race) in the 12 Massachusetts communities where less than half the population is white

Share of Population Vote Share
City/Town White Hispanic Black Asian Clinton Warren Diff.
Aquinnah* 40% 0% 0% 1% 82% 86% +4%
Everett 46 23 19 6 67 69 +2
Malden 47 9 16 24 70 72 +2
Lowell 49 20 7 21 64 65 +1
Springfield 33 44 19 2 74 75 +1
Lynn 38 39 12 8 67 67 +1
Randolph 36 8 38 12 75 76 +1
Boston 45 19 23 9 81 81 0
Holyoke 43 51 3 2 70 70 0
Brockton 37 11 39 2 71 71 0
Chelsea 22 66 6 3 79 79 0
Lawrence 16 79 2 2 82 81 -1
Statewide 73 11 7 6 60 60 0

*42 percent of Aquinnah residents are Native American.

Sources: 2013-2017 American Community Survey, Massachusetts secretary of the commonwealth

We can take a closer look at how Warren performed among Hispanic, black and Asian voters specifically by zooming in to the precinct level in Boston, whose neighborhoods remain fairly segregated by race.12 This data reveals that Warren may hold special appeal among nonwhite voters after all — specifically, Hispanic voters. Warren did between 1 and 6 points better than Clinton in all 10 Boston precincts where, as of 2010, at least 50 percent of residents were Hispanic. At the same time, the precinct data also seems to confirm that black voters are truly agnostic about Warren. On average, Warren did only 1 point better than Clinton in Boston’s 56 majority-black precincts, with very little precinct-by-precinct deviation. It was hard to arrive at a conclusion about Asian voters; Boston had no majority-Asian precincts in 2010. The three precincts where Asians constituted a plurality of the population differed dramatically from one another. In one, Warren did 2 points better than Clinton; in the other two, Warren did 4 and 10 points worse.13

People of color cast about 40 percent of all votes in the 2016 Democratic primary, so anyone who hopes to be the party’s 2020 nominee must win a healthy share of nonwhite voters. On the plus side for Warren, it’s a good sign that her 2018 performance among these voters wasn’t actively bad, given that Sanders (whose natural constituencies overlap with Warren’s) struggled so much to win minority voters in the 2016 primary; indeed, her campaign managers ought to find her recent performance with Hispanic voters downright encouraging. But on the other hand, Warren has also shown no particular knack for connecting with black and Asian voters. She has plenty of other strengths as a candidate — including her ideology and fundraising ability — but fostering more of a base among nonwhite voters could give her the complete package.


  1. The fact that I was born and raised there certainly has nothing to do with it.

  2. She formed an exploratory committee but has not yet officially announced her candidacy.

  3. Each city or town got a score between 0 and 100, which represented the average of its statewide percentile ranking for median household income and its statewide percentile ranking for percentage of the population over age 25 who possess a bachelor’s degree or higher. All city and town data (excluding precinct-level data in Boston) in this article are five-year estimates from the U.S. Census Bureau’s 2013-2017 American Community Survey.

  4. The other is in Charlestown.

  5. All the towns in the table are at least 89 percent non-Hispanic white.

  6. By median age.

  7. Gordon College, a Christian school of about 2,000 students.

  8. Home to Amherst College, Hampshire College and the main campus of the University of Massachusetts.

  9. The location of Harvard, MIT and Lesley College.

  10. Where you’ll find the Purple Cows of Williams College.

  11. The 20 cities and towns where Warren outperformed Clinton most dramatically were an average of 94 percent non-Hispanic white; the 20 cities and towns where Warren fell the furthest short of Clinton were an average of 90 percent non-Hispanic white. In both cases, that’s significantly higher than the statewide average of 73 percent non-Hispanic white.

  12. Unlike for cities and towns, the most recent demographic data for Boston precincts is from the 2010 Census, so it’s a bit out of date. But because Boston hasn’t redrawn its precincts in years, the only forces changing the population makeup in these areas are natural demographic shifts, which tend to move relatively slowly, so this data can still give us a decent idea of who’s voting where in Boston.

  13. The first precinct includes Chinatown, a neighborhood that is undergoing gentrification, which may have helped Warren. The other two precincts were in Dorchester.

Nathaniel Rakich is FiveThirtyEight’s elections analyst.