In the immediate aftermath of the November election I re-examined my predictions for the Presidential Election in Pennsylvania using the data available at the time. Subsequently the official election results have been published by the Pennsylvania Department of State along with the final registration numbers for November 8, 2016, allowing me to revisit the question of how Trump managed to win what many, including myself, thought was an increasingly unwinnable state for the Republicans. Trump did win, and the short answer is he got people to turn out in higher numbers than usual.
Pennsylvania Final Results
Bizarrely, the official numbers of the state of Pennsylvania do not list write-ins or even the votes of Evan McMullin, so I am obliged to seek the data from Dave Leip’s Atlas of US Presidential Elections, where the totals are more thoroughly listed. Trump won Pennsylvania in the final tally by 44,292 votes, a smaller margin than originally stated: the final vote count was Trump 2,970,733 (48.17%) to Clinton 2,926,441 (47.46%), with write-ins and various third party candidates taking a not inconsequential 269,524 votes (4.37%). Write-ins alone comprised 48,259 votes, a larger total than the margin between the two main candidates. As it is hard to know how many of the third party voters would normally have voted for the Republican or the Democratic candidate, it is foolish to try to speculate or cast blame on these independent voters. Instead we need to look at overall turnout and the places Trump increased the Republican vote total. Overall 6,166,698 votes were cast, which represents an increase of over 150,000 votes cast in the previous high turnout, the 6,010,519 voters who turned up in 2008. In that election a record number of registered voters was also established, with 8,755,588 registered voters, a number not quite matched by the 8,722,977 registered voters on November 8, 2016. Thus turnout in the 2016 election was 70.69% compared with 68.65% in 2008. The previous high turnout was in 1992, when an astonishing 82.80% of voters turned up on election day, though the actual number of registered voters, at slightly under 6 million, was only 75% of the most recent figures. However, I will return to the 1992 election as I see a parallel in the two elections that I believe explains much of the data I will present.
Trump’s total exceeded by 176,886 votes the previous Republican high of 2,793,847 posted in 2004 by George Bush. Clinton’s total was the lowest by a Democrat since Al Gore in 2000, and was 12,000 votes fewer than John Kerry received in 2004 when he defeated Bush by 145,000 votes. Obama received 2,990,274 votes in 2012 and an all-time Pennsylvania record 3,276,363 votes in 2008. Obama’s 2012 total would have been sufficient to defeat Trump, indicating that Clinton lost the support of more than 60,000 Obama supporters. Since it is obvious that new voters have come into the process and some have died or moved or not voted in the interim, it is likely that Clinton gained many new voters as well; thus she is likely to have lost far more than 60,000 Obama voters, perhaps well over 100,000, although this is too complicated to sort out precisely.
What is clear is that Clinton’s vote was down somewhat from Democratic norms of the past four elections while Trump managed to dramatically increase his vote compared to recent Republicans, despite the fact that the demographic data point towards the increasing likelihood of a Democratic Party favorable population both now and in the near future. So what happened?
Philadelphia
The traditional view is that in Pennsylvania, Democrats run up the score in Philadelphia and then hang on in the rest of the state to win. Obviously this did not work as planned so any investigation of voting results in Pennsylvania must begin in the state’s largest city. Philadelphia’s results were not particularly abnormal: Clinton received 82.30% of the vote in the city to Trump’s 15.32%. Her vote total of 584,025 was down about 5,000 votes from Obama’s 2012 total of 588,806 votes, while Trump gained some voters for the Republicans in Philadelphia for the first time since 2004 receiving 108,748 votes, 12,000 more than Romney but still 9,000 votes shy of McCain’s total of 117,221 votes in 2008. In the end Clinton ran up a respectable 475,277 vote margin of victory in Philadelphia, with a total turnout of 709,618 besting 2012 by almost 20,000 votes but failing to equal the high of 717,329 votes cast in 2008. Effectively Clinton got the job done in Philadelphia, building a margin larger than Kerry in 2004 (412,106), but slightly under-performing Obama’s total from 2008 (478,759) as well as his 2012 margin (492,339). I had stated that Clinton needed and was likely to have a margin of between 460,000 and 500,000 and she hit that mark right in the middle.
The catch is that the Philadelphia vote in the past three elections made up between 11.68% (2004) and 12.00% (2012) of the total Pennsylvania vote; however in this election the total of 709,618 votes cast, respectable as it was, was only 11.51% of the total number of 6,166,698 votes cast in the state. Thus, Clinton would need to over-perform in the remaining 66 counties in Pennsylvania to make up the gap; she did not, receiving 2,342,416 votes in the rest of the state, the lowest Democratic total in the last four elections (Obama 2,401,468 in 2012, and 2,680,383 in 2008, and Kerry 2,395,890 in 2004). Trump, on the other hand, got 2,861,985 votes outside of Philadelphia, 200,000 votes more than Bush’s high of 2004 (2,663,748) and far higher than McCain (2,538,664) in 2008 or Romney in 2012 (2,583,967). This despite the fact that the ‘home counties’ near Philadelphia are increasingly Democratic in their vote and are the major population growth area in the state. I will take a closer look at these counties before moving on to other part of the state.
One final note of interest: Total voter registration in Philadelphia was 1,102,560 a number higher than all but the 2008 election when 1,126,760 voters registered. The 1.1 million registered voters in Philadelphia represents 12.64% of all registered voters in Pennsylvania in the 2016 election, the lowest percentage of the last four elections (12.92% in 2012, 12.87% in 2008, 12.70% in 2004) despite the fact that the population of Philadelphia increased by over 40,000 people in the past five years, over 40% of the total increase in a state where the vast majority of counties saw population decline at the same time. If ‘illegitimate’ voting were to occur in Philadelphia, one would expect a higher turnout than normal and a higher number of registered voters than normal, neither of which is the case. In fact it is the EXACT opposite! Philadelphia’s population represents 12.24% of the state of Pennsylvania, yet only 11.51% of all votes cast in Pennsylvania came from Philadelphia. Voter turnout was 64.36% compared to 70.69% for Pennsylvania as a whole. Had the turnout in Philadelphia merely matched overall turnout in Pennsylvania (i.e 70.69% of registered voters, not even total residents) then we would expect to have seen 779,400 votes cast, 70,000 more than were actually cast. If the extra votes broke as the rest of the vote in Philadelphia did, then Clinton would have received 57,431 more votes, while Trump would have gained an additional 10,691 votes. The final totals in Pennsylvania, keeping everything else constant would have been Clinton 2,983,872 votes to Trump 2,981,424 resulting in a very narrow 2,448 vote victory for Hillary Clinton in the state of Pennsylvania, a number for Clinton I had actually expected her to receive.
Incidentally, 45.27% of the total 2015 population (US Census data) of Philadelphia voted, compared to 48.17% of the total 2015 population of Pennsylvania (which includes the data from Philadelphia!). If anything the results in the city of Philadelphia and other major cities in the country indicate that the urban vote is dramatically underrepresented and that rather than look for voter fraud in these counties, one should seek reasons the vote totals are not higher than they are. Here again I reiterate my claim that Donald Trump is a fucking liar!
Philadelphia Suburbs (Montgomery, Delaware, Bucks, and Chester Counties).
Things were brighter for Clinton when the entire five county metropolitan area is taken into account.
Year D Vote R Vote D Margin % of PA total Vote 4 County margin
2016 1,326,251 662,621 +663,630 33.50% +188,353
2012 1,278,876 663,120 +615,756 34.26% +123,417
2008 1,345,107 662,715 +682,392 33.85% +203,633
2004 1,200,000 700,770 +499,230 33.16% +87,124
Prediction below
2016 (Low) 1,224,000 693,000 +531,000 33.25% +71,000
2016 (exp) 1,283,000 650,000 +633,000 33.45% +153,000
2016 (High) 1,300,000 651,000 +649,000 34.00% +175,000
Clinton not only performed well in the five counties above, she exceeded my best case scenario by 26,000 votes, while Trump received the lowest vote total of any Republican in the last four elections in the five counties, and was about 10,000 votes above the vote total I expected him to get. The overall margin between Trump and Clinton thus stood at 663,630 votes, the second largest gap in the last four elections, behind Obama’s romp in 2008. Again, for the most part these counties behaved as expected in turnout and in terms of voting preference. Overall turnout was 2,067,646 voters, which made up 33.53% of the total vote in Pennsylvania. The five counties population in 2015 totaled 4,093,906 residents, 31.98% of the population of Pennsylvania, while voter registration was 2,908,123 which represented 33.34% of the total registered voters in the state. Turnout for the five counties collectively was 71.03%, slightly higher than the turnout for the state (70.69%).
Thus Clinton over-performed in the counties surrounding Philadelphia both in terms of total votes received as well as in the margin of victory over Trump. Overall the vote total in the five counties was the highest ever, and at 33.50% of the total vote in the range of previous elections (34.08% in 2012, 33.67% in 2008, 33.08% in 2004). Again, it bears repeating that in the five counties of the Philadelphia area, Clinton performed as expected and perhaps slightly better than expected. The increase in voter turnout reflects the population increase between 2010 and 2015 in these counties, which totaled 84,896 in a state that increased by a total of 100,124; thus 85 of every 100 new residents of Pennsylvania come from the five counties. The entire rest of the state had a net growth of 15,000 residents in the same period.
And yet the remaining 62 counties in Pennsylvania increased their share of the total turnout from 3,792,590 votes in 2012 to 4,101,052 votes in 2016, a change of more than 308,462 voters showing up in 2016! the 4.1 million voters casting ballots in the 62 counties referred to above was also 114,563 votes more than the record 3,986,489 voters turning out in 2008 in these same counties. Is this plausible or should we be suspicious? It is plausible, if one takes a few factors into account. First of all, the five counties represent 31.98% of the total population and yet accounted for 33.50% of the vote in 2016; in other words they vote in higher numbers than in the typical county in the state as a whole, led by the astonishing 77.02% turnout of Chester County, which also has the highest percentage of college graduates of any county in the state at 49.1% (the state average is 28.6%). The short answer is that the rest of the state turned out at a higher rate than is normal. Sean Trende suggested as much in a 2013 article: namely, that white voters were sitting on their hands in 2012 and perhaps even going back to the 1990s.
Which brings me back to 1992. In this year Bill Clinton managed to defeat George H.W. Bush in large part because many voters cast ballots for the third party candidate Ross Perot, who promised to bring back jobs and railed against NAFTA. How can I say it was because of Perot? Here are the vote totals for the Democratic candidate for the four elections from 1984 to 1996 in order: 2,228,131; 2,194,944; 2,239,164; 2,215,819. A mere 44,220 votes separates the highest total from the lowest total and yet Mondale lost by 7.35% to Reagan in 1984, Dukakis lost by 2.31% to Bush Senior in 1988, but Clinton won by 9.02% in 1992 and again by 4.17% in 1996. No Democrat earned 50% of the vote in Pennsylvania until Gore in 2000, who received 2,485,967 votes, a 250,000 vote improvement on previous Democratic candidates.
If we turn to the Republicans, Reagan pulled in 2,584,323 votes in 1984, a number unmatched until Bush reached 2,793,847 votes in 2004 twenty years later, when he lost to Kerry by 140,000 votes. From 1984 the numbers declined steadily: Bush Senior earned 2,300,087 votes in 1988, but only received 1,791,841 votes in 1992, a drop of half a million votes. Where did those votes go? Not to the Democrat Bill Clinton, who earned only 11,000 more votes than Walter Mondale in the rout of 1984. Instead, Ross Perot picked up 902,667 votes which, added to Bush’s total, brings the not-Clinton vote to 2,694,508 votes, 100,000 more than Reagan received in 1984. Perot’s numbers dropped dramatically when he ran a second time to a still respectable 430,984 votes. Bob Dole’s vote total however was a mere 10,000 votes more (1,801,169) than Bush in 1992, while Clinton’s vote total declined by 24,000 votes from his 1992 total. Thus, nearly a half a million voters stayed home in 1996 (4,501,307 votes cast) compared to 1992 (4,959,810 votes cast) a number similar to the total votes cast in 1988 (4,536,251).
It is my belief that in Trump we have merged the Perot and the classic Republican voter, in a similar vein to what are often called Reagan Republicans. In 1984, Reagan could trounce Mondale but in 2016, a weak Democratic candidate such as Hillary Clinton, who received the fewest votes of any Democrat since Al Gore in the 2000 election (a time which seems like ancient history in the technological era of get out the vote strategies), could almost defeat the combined forces of the typical Republican voter added to the once in a while Perot/Reagan Democrat voter. This is mainly owing to demographic changes in the state in the intervening 30 years.
So, I end this article with a positive spin. It is apparent to me that this ‘surge’ of voters is essentially a one time force, a unique combination of anger, age, demography, and unusual candidates. If form holds, one would anticipate the Democratic vote in 2020 to stay relatively steady, while the Republican vote would seem to be to be about to crash back to levels of previous elections, somewhere in the 2.6-2.8 million vote range as older voters, who preferentially chose Trump, die, as younger less-white voters join the voting ranks and most especially, as voters realize they have been suckered and drop out of the system again, much as they did in 1996.
In the next article I will look at the remaining 62 counties in Pennsylvania in detail and try to parse the data to discover where this ‘surge’ of voters for Trump came from, who they are, and analyze the likelihood of this result repeating itself in four years time. (hint: NO!)