The Economist - USA (2020-11-13)

Sources:USCensusBureau;DecisionDeskHQ;PewResearchCentre;TheEconomist *Countieswithover95%ofvotesreported

Predictedchangeinmarginofvictory:+4Dem Predictedchangeinmarginofvictory:+4Rep Predictedchangeinmarginofvictory:+10Dem

2016 +17Dem 2020 +21Dem 2016 +38Dem 2020 +34Dem 2016 +36Rep 2020 +26Rep

Actualchangeinmarginofvictory Actualchangeinmarginofvictory Actualchangeinmarginofvictory

Whitecollege-educatedvoters,% Hispanicvoters,% Whitevoterswithoutdegrees,%

-40

-30

-20

-10

0

10

20

0 20 40 60 0 20 40 60 80 100

-40

-30

-20

-10

0

10

20

30

0 20 40 60 80 100

→PollsexpectedwhiteswithoutcollegedegreestodriftbacktowardstheDemocrats.TheystuckwithDonaldTrump

UnitedStatespresidentialelections2016-20,changeinDemocratic

votemarginpredictedbyPewv countyresults*,percentagepoints

Whitecollege-educatedvoters Hispanicvoters Whitevoterswithoutdegrees

Totalvotes,m

0.5

1.0

2.0

↑MoreDemocraticthanin 2016 ↓A lotmoreRepublicanthanin 2016 ↓SlightlymoreRepublicanthanin 2016

Miami-Dade,

Florida

Indianapolis

suburbs

Laredo,

Texas

TheEconomistNovember 14th 2020 77

F

or thesecond presidential election in a

row, polls underestimated support for

Donald Trump and the Republicans. In

states that have mostly finished counting

votes, the error of an average of presiden-

tial polls released during the final two

weeks of the campaign was 5.5 percentage

points, nearly double the average miss of

3.1 percentage points registered in 2000-16.

Moreover, whereas both major candidates

in 2016 benefited from state-level errors—

Hillary Clinton won California by seven

points more than pollsters expected—this

year, virtually all of the misfires were un-

derestimating support for Mr Trump.

What went wrong? Exit polls do not pro-

vide a trustworthy answer: their estimates

of vote margins within each demographic

category, as well as the share of the elector-

ate each group represents, are often biased

and differ vastly from other sources. The fi-

nal analysis will have to wait until all bal-

lots are counted and rigorous post-election

studies can be conducted. However, statis-

ticians can unearth promising clues by ex-

ploring the relationships between official

vote totals in counties where tabulation is

nearly complete and those areas’ demo-

graphic makeup. These data show that Mr

Biden failed to realise one of his central

electoral promises: clawing back some of

Democrats’ recent losses among white vot-

ers without college degrees.

Among the most reliable analyses of the

election of 2016 is a study published by the

Pew Research Centre, a polling and re-

search organisation, which surveyed 3,000

voters confirmed to have cast ballots that

year. Pew also published a nationwide poll

this October, making possible direct com-

parisons of how voting intentions had

changed within each demographic group.

Pew found that Mr Trump had made in-

roads with black and Hispanic voters, trim-

ming his deficits with these groups by four

percentage points, while suffering an off-

setting four-point decline among college-

educated whites. However, most of Mr Bi-

den’s predicted gains relative to Mrs Clin-

ton came from whites without degrees: he

was expected to reduce Mr Trump’s margin

with such voters by ten percentage points.

The county-level data affirm some of

these findings. Places with lots of college-

educated whites, particularly in suburbs,

did indeed swing towards Mr Biden this

year—a trend that will probably make him

the first Democrat to win Georgia or Arizo-

na since Bill Clinton. Similarly, Mr Trump

fared far better in majority-Hispanic coun-

ties than he did in 2016, padding his advan-

tages over Mr Biden in Florida and Texas.

However, there was scant evidence that

white voters without degrees preferred Mr

Biden to Mrs Clinton. Mr Trump’s margins

of victory in white working-class counties

this year were just as large on average, and

in some places even bigger, than in 2016. As

a result, Mr Biden had to rely mostly on his

strength in affluent suburbs to rebuild

Democrats’ “blue wall” of Wisconsin,

Michigan and Pennsylvania. He won these

states by far smaller margins than pollsters

expected, and just barely squeaked by in

Wisconsin, the decisive state in the elec-

toral college in both 2016 and 2020.

So far, most attention has focused on Mr

Trump’s gains among Hispanics, which ap-

pear to have been even greater than polls

foresaw. However, the durability of his

edge among less-educated whites, a much

larger group than Hispanics, is far more

electorally consequential. No one knows

how much of this affinity is specific to Mr

Trump, and how much will carry over to

other Republican candidates. The party’s

electoral future depends largely on its abil-

ity to keep these supporters, while making

itself more palatable to the college-educat-

ed white voters who have abandoned it. 7

Once more, less-educated whites

spurned the Democratic nominee

Déjà vu all

over again

Graphic detailPolling in America