“Every vote counts” has never been more emblematic in any democratic election than the 2016 United States General Election. President-elect Donald Trump won an estimated 306 electoral votes en route to his victory. Meanwhile, Democratic candidate Hillary Clinton won the popular vote, but lost key battles in several pivotal states. The Reckoner presents a statistical breakdown of the 2016 American General Election.

Swing States Breakdown

This year, Trump managed to sweep the five crucial states of Florida, Michigan, Wisconsin, Pennsylvania, and North Carolina, but he won by extremely slim margins ranging from 0.3% to 3.8%. In the last election, Democratic candidate Barack Obama won all but one of these states. What changed? The following is a breakdown of Donald Trump’s and Hillary Clinton’s vote shares, in addition to a county map comparison between Barack Obama’s 2012 victory (left) and the 2016 election (right).





As in most other states, Trump received the overwhelming majority of Florida’s white vote. He was able to win 62% of votes from college graduates, an unanticipated 10% increase from Romney’s 2012 share. As expected, Clinton strongly appealed to minorities in Florida, where she won 84% of black voters and 62% of Latino votes. However, considering that black and Latino voters only accounted for 14% and 18% of the Floridian electorate respectively, Clinton’s advantage did little to negate her lack of appeal to white voters.




The trend of Clinton’s low favourability amongst white voters, especially those lacking college degrees, continued in Michigan. Moreover, her widespread popularity among minorities was also apparent. Contrary to recent surveys of college graduates in America, which found that college graduates are more left-leaning, 50% of Michigan’s voting population with graduate degrees elected Trump. Meanwhile, Clinton received only 45% of the vote from the same category. This 10:9 ratio was apparent for all other groupings based on education, except in voters with post-graduate degrees, who voted 61% in favour of Clinton. Of course, the composition of the electorate with a post-graduate education was relatively small, at 33%. When considering income levels, Trump performed well with voters who had annual earnings exceeding $30 000. In key states, voter income was not a strong indicator of a voter’s preferred candidate, especially when compared to factors such as race and education.




In Wisconsin, Clinton outperformed Trump among voting college graduates by a margin of 4%, one of the few key states where she had surpassed Trump in this category. Despite only making up 28.4% of eligible voters, citizens with a bachelor’s degree or higher were actually overrepresented, as they made up 42% of voters on election day. Trump again found success in 86% of Wisconsin’s white electorate, with 53% of them voting for the Republican nominee. Trump won almost every income group except for those who earned between $30 000 to $49 000 annually, where he tied with Clinton.




The voting patterns of age groups among key states were well demonstrated in Pennsylvania: voters under the age of 45 favoured Clinton, while older voters favoured Trump, both by margins of almost 10%. The division among race categories was also exemplified in this state, where Clinton received 92% of the black vote and Trump received 56% of the white vote. While this was anticipated due to what had already been established on the campaign trail, the divide was more prominent. Most voters without post-graduate degrees once again elected Trump. Clinton performed unexpectedly well among voters who annually earned less than $50 000, especially given her performance in many other states, within this category.


North Carolina


Of all of the outlined key states, Trump won by the largest margin in North Carolina (3.8%). His popularity among white voters placed him ahead of Clinton once more, despite being vastly unpopular among black voters, who gave him only 8% of their vote. While Trump did not win the Latino vote in any of the states on the list, he performed considerably better than media outlets had predicted. His appeal to those with annual incomes below $50 000 was less than in other key states, but his popularity with voters annually earning above $100 000 tipped the scales in his favour.

Polling Error

People were shocked when Donald Trump defied every pollster in the country, picking up votes left and right in the all-important swing states. Where did all these voters come from? The most popular theory revolves around Trump’s “silent” support in the rural counties, where voters either did not want to reveal their preferred candidate or were not polled in the first place. How legitimate is that claim? The theory was put to the test.


It was difficult to isolate rural voters from the rest of the American people, so some generalizations were made. Considering that one of Trump’s biggest voting margins in this year’s election was with the white and uneducated group, criterion was used specifically for this relation. R2, or coefficient of determination, had a value of 0.6656, or around 67%. This was the percentage of variation that variable x could predict in variable y. The R-value, or the coefficient of correlation, was around 0.81. Both of these values conveyed the strength of the relationship between the x and y points. Generally, the higher the R and R2 squared values, the more correlated the data sets are. The graph showed a steady linear relationship between the two factors. 

Given R and R2, it was determined that there existed a strong correlation. But why?

Many would point to a strong rural turnout, which could be true, given that the greatest percentages of polling error were in Republican states such as West Virginia and Idaho. However, Donald Trump received fewer total votes than Mitt Romney did in 2012. It was much more likely, then, that either Democratic voters chose to vote for Trump, or those polled simply lied and said they would vote Clinton, as they were afraid of public backlash.

On the other end of the spectrum, Trump also did worse than the polls suggested in most blue states, such as New York, Hawaii, and District of Columbia. This could have been because of extreme distaste for Trump, or even something as simple as complacency by pollsters in these areas. Nevertheless, looking at the standard deviation of 3.23 and the mean value of 1.82 from the y points, it was determined that most states had polls which overrepresented Clinton.

However, the majority of polling errors greater than 5% in either direction did not really affect the presidential race, since those state elections ended in landslides regardless. The most important states to look at only swung by 1% to 3%–namely Wisconsin, Michigan, and Pennsylvania. Wisconsin had a polling error of 3%, while Michigan and Pennsylvania had polling errors of 2%. Pollsters originally believed these states would be close but safe for Clinton, so a polling error of 2% to 3% essentially meant that each state was decided by a couple thousand votes. Every other election, the rural county votes would be cancelled out by the huge cities in Milwaukee, Pittsburgh, Philadelphia, and Detroit. This time, however, the rural votes overwhelmed the major commercial centres, allowing Trump to sweep all three states.  

Of course, these polling errors did not necessarily only occur in rural areas. Perhaps the 2% vote swing appeared in city centres or suburbs. The data was inconclusive, due to the nature of its generalizations, but the voting patterns of key states for Trump suggested that there was an influx of voters from the rural counties this election.

Qualitative Studies

Generally, election polls give reasonably accurate estimations of the votes a candidate will receive, based on quantitative survey data. However this year’s election showed that perhaps polls ignore the more “human” qualities of a voter.  Polls are fragile because they do not reflect the ideas and thought processes of the average voter. These “why” questions must be asked through a conversation. In an online survey platform known as Remesh, a conversational study that looked into 5 attributes of each candidate was conducted with 275 likely voters.



Source: Team Remesh

As shown in the graph above, it is clear that an astounding amount of voters across the political spectrum were unhappy with their choices in this year’s election. Even national exit polls, which were conducted with many more voters, reflected this point.


Source: New York Times


Both Clinton and Trump received more than 50% in unfavorability, and a staggering number of voters said they chose one of the two candidates simply because “the other was worse.” Additionally, Clinton was rated higher than Trump on four of five attributes, most notably in the category of ability. It seemed that half of the American population was willing to have a more direct, rather than a more experienced, candidate. This showed that some Americans seemed to be voting against the establishment, and highly valued the notion of a transparent government, while others did not see these as major issues in the political system.

Cheques and Handshakes

America’s Electoral College system allocates 538 electoral votes between all 50 states and the District of Columbia. The system distributes electoral votes to each state based on population, meaning some states receive more than others. The winner-takes-all in every state, meaning that if a candidate receives 50.1% of the popular vote in a state, then they receive all of its electoral votes. This results in swing states such as Florida, Pennsylvania, and North Carolina receiving more concentrated efforts from candidates than others, because they have a large number of electoral votes and do not reliably vote for either major party. So how did the candidates distribute their time and money in the crucial stages of the race?


The Spending Game

Hillary Clinton vastly outspent Donald Trump across the board. In every swing state, her team spent millions more on advertising, yet she lost the majority of these races, showing that the argument “money buys elections” may be untrue. It is interesting to note that Clinton invested in regions such as Georgia, Nebraska, and Arizona, which did not seem very favourable for her. Ultimately, these states voted against her by a relatively large margin. However, Clinton neglected to spend any money on Wisconsin, a Rust Belt state that heavily contributed to her loss.


Heat maps of Trump’s and Clinton’s campaign spending in the election. Only states with more than one million dollars in spending are shown.

State Visits


Campaign visits by each candidate in the 3 months leading up to Election Day.

From 1 August 2016 to Election Day, Trump made over one and a half times as many campaign stops as Clinton. And while Clinton mostly stuck to traditional swing states like Florida and Pennsylvania, Trump made numerous visits to traditionally democratic states like Maine and Virginia. What is also important to keep in mind is that Trump’s campaign stops were much larger gatherings when compared to Clinton’s.  Although the effect of these stops on the average voter is debatable, it is indisputable that Trump made himself much more visible and accessible to the public in the months leading up to the election.


The majority of Americans disliked both candidates, but immense support from the rural counties helped Trump sweep the swing states. During their campaigns, Clinton spent more on advertising compared to Trump, while Trump made more visits than Clinton. Polls underrepresented Trump supporters for many key areas, surprising everyone when 306 electoral votes turned red.



New York Times

Kaiser Family Foundation

The Associated Press

The Times

United States Census: Fact Finder



United States Federal Election Committee

270 To Win

The Guardian

Team Remesh