While Hillary Clinton remains the heavy favourite in the U.S. presidential election, recent polls suggest her victory isn't looking as assured as it once did.

Recently, the CBC's Presidential Poll Tracker was projecting that enough states were considered "safe" for the Democratic nominee to secure victory even if she lost all of the remaining swing states to Donald Trump. If an election had been held last week, a Clinton win would have been projected with more than 95 per cent confidence. 

That's no longer the case because Clinton's electoral college vote tally among safe Democratic states has dropped from 273 — just above the 270 needed to take the White House — to 253. Her lead over Trump in the national polls has slipped from a high of 6.4 points among decided voters in early August to 5.1 points today.

This movement in the electoral college has largely been driven by Clinton's narrowing lead in the national polls, though some surveys at the state level also point to a few tightening races. Still, one poll shows that a linchpin state in Donald Trump's electoral map may be moving out of his reach.

The three most recent U.S. polls in the Presidential Poll Tracker point to a trend line that is slowly moving against Hillary Clinton.

  • An Ipsos/Reuters poll put Clinton ahead by five points among decided voters. Her edge in the Ipsos poll done a week earlier was six points.
  • A poll from Morning Consult had Clinton ahead by four points among decided voters, compared to a lead of seven points earlier in the month.
  • And the daily tracking poll from UPI/CVOTER has gone from a Clinton lead that was as wide as seven points to a tie with Trump.

These polls moved Pennsylvania and its 20 electoral college votes from "safe" to "likely" Democrat status, as the last survey conducted in Pennsylvania dates from Aug. 7. That coincided with Clinton's widest lead in the national polls. The state-level data is now relatively older, so the negative trend line in the national numbers carries increased weight.

Nevertheless, Clinton is still projected to be ahead in Pennsylvania by about eight points.

Some swing states inching to the right?

A handful of swing states have seen polling numbers that are more positive for Trump than has been the case for some time. But it's difficult to discern a trend line from them.

Gravis Marketing, a polling firm that has generally reported better numbers for Trump, suggests the Republican nominee's chances are looking up in the Carolinas.

In South Carolina, Trump was up by five points among decided voters, whereas a Public Policy Polling (PPP) survey earlier this month had him ahead by only two points.

USA-ELECTION/TRUMP

Donald Trump, addressing supporters at a recent rally, questioned whether rival Hillary Clinton 'looks presidential.' (Jonathan Ernst/Reuters)

In North Carolina, Trump was up by one point among decided voters. Recent polls from Marist College and PPP had Clinton ahead in the state by nine and two points, respectively.

And in Iowa, a CBS/YouGov poll put Clinton and Trump in a tie. This is in line with what other surveys have been showing in what appears to be the most tightly contested state in the country.

But is this a sign of an improving — though still very challenging — electoral map for Donald Trump? These pollsters haven't been active in these three states until recently, so it's impossible to know for certain whether their trend line is leaning towards the Republicans. Nevertheless, it does match the national trend.

Bad news for Trump in Ohio

It should still be noted that Hillary Clinton remains in an enviable position. Though her tally in the electoral college is 253 votes among "safe" states only, that extends to 320 votes when "likely" states are included. This means Clinton could lose battleground states like North Carolina, Nevada and Iowa and still have some 50 electoral college votes to spare.

She could then lose any two of Florida, Ohio and Pennsylvania and still take the White House.

State breakdown, U.S. projections, Aug. 23

U.S. presidential election projections by state, as of Aug. 23, 2016. Size of circle represents the number of votes the state is worth in the electoral college.

Ohio is particularly important to Donald Trump. It was host to the Republican National Convention in July. No Republican has won the presidency without Ohio. 

That makes the YouGov poll particularly problematic for him. Though two recent surveys (from Marist and Quinnipiac) put Clinton up two to four points, the YouGov survey has her ahead by six points among decided voters.

While 55 per cent of Ohioans said Clinton "is prepared to be commander-in-chief," just 35 per cent said the same thing about Trump. And while 51 per cent felt Clinton "is a risky choice," fully 70 per cent thought so for Trump.

Nevertheless, trust remains Clinton's Achilles heel — recent news about more unreleased emails certainly won't help. The poll found just 40 per cent of Ohioans felt Donald Trump "tells the truth," but only 33 per cent said the same thing of the Democratic nominee. 

This does limit Clinton's ability to pull ahead more definitively.

In Iowa, where the YouGov poll suggests there's a tie and Trump has just as much support as he does in Ohio, he scored worse on all of these measures. But so did Clinton. That suggests Trump may have a floor that his campaign's problems can't sink him below, but Clinton can be moved into a tie with Trump in some states in part due to her problems on issues like trust.

It raises the question: just how badly would Donald Trump be losing if his opponent wasn't Hillary Clinton?


The Presidential Poll Tracker includes all published mainstream surveys, a list of which can be found here. The polls are weighted by sample size and date, as well as the reliability of each pollster as rated by FiveThirtyEight.com. The electoral college is projected by applying the same weighting standards to state-level polls and combining this with a uniform swing model, based on how the current national polling average compares with the 2012 presidential election. Surveys included in the model vary in terms of sample size and methodology and have not been individually verified by the CBC. A full methodological explanation can be found here.