Look, let me begin by stating that, although I have never personally visited any of the 54 counties that I am about to discuss, I am sure that they are not horrible places. Not only are they likely (having not visited any of them) quite varied and interesting, some are probably very beautiful and I am sure many of the people in most of these counties are pleasant and well-meaning. I merely structured my essays (and the title) to reflect the transition from places where the Democrats were successful like Philadelphia (in Part I), to places where the successes were less overwhelming, like Bucks County, to places like Lackawanna County where Clinton’s victory was more like a defeat, to address in this final entry the many counties in Pennsylvania in which Trump was victorious.
As I am not a fan of Donald Trump, this outcome is, in my view, a disaster for America, and I now think of the places that voted for Trump as somehow less generous places than places that did not support him. Nothing personal, but if people support an egomaniacal, lying, devious,sleazy, perverted, and possibly insane billionaire in the misguided notion that he and his cronies are, in any way shape or form, going to ‘make America great again’ (whatever that means!) they have clearly gone off the rails somewhere and so it is my goal to figure out how (that is, how statistically was it possible for Trump to win) and why (who voted for Trump and what motivation besides pure evil could be behind their dubious decision).
If that sounds snobby and elitist so be it. I happen to care about my country and obviously I think about the long term damage a period like this will have on a country I increasingly worry is in a state of terminal decline. I genuinely worry we are a giant Argentina, Turkey, or even Russia in the making, where rule of law is seen as a choice and not a requirement, where facts can be ignored or made to disappear, where might makes right. If I can figure out, by studying data and seeking patterns and trends that might lead to a way this fate can be avoided I want to get the news out there.
So, where were we? Oh yes, Clinton won 11 counties of the 67 counties that make up Pennsylvania. I examined two other large counties relatively close to Philadelphia, Lancaster and Berks County in previous entries as well. The total vote in these 13 counties comprised 60% of the total votes cast in the 2016 election in the state, and Hillary Clinton had built an advantage over Donald Trump that was larger than the margin Barack Obama had amassed over Mitt Romney in 2012. Yet Clinton lost by 44,292 votes to Trump in Pennsylvania, in large part because in the remaining counties in Pennsylvania (all carried by Trump) voter turnout was up over 8% (8.26%) compared to the turnout in the 2012 election, while turnout in the 13 counties already discussed, the counties which contained almost all the new population growth in Pennsylvania in recent years, saw only a 6.41% increase in turnout.
How to explain the surge in voter turnout in these 54 counties, which overall saw a steady decline in population in recent years relative to the counties already discussed? Eleven of the thirteen counties I have already discussed increased in population between 2010 and 2015, while only two had population declines. The total increase in population in these thirteen counties was over 130,000 net new residents, which is 30,000 more people than the growth of the entire state in the same period. Thus, the remaining 54 counties clearly lost 30,000 residents in the same period and yet the total number of votes cast in these counties increased by 188,136. Clearly something interesting happened here that needs further examination.
Overall 23 counties saw a population increase between 2010 and 2015. The 12 counties not yet discussed that increased in population had a total net increase of about 35,000 people; therefore the remaining 44 counties lost over 65,000 people in a five year period, led by Westmoreland County near Pittsburgh, which lost 7,213 people in five years and over 12,000 people since the 2000 Census. Westmoreland County also had the highest turnout ever for a presidential election in 2016. Should we be worried about illegal voting? Most likely no, but the reason for such a high turnout needs to be explored in detail.
The Trump-supporting counties come in many flavors so it will be necessary to break them down into categories in order to be able to make sense of individual differences in voting patterns. First there are the counties that were won by Obama in 2012 which Clinton lost: Luzerne, Northampton, and Erie Counties. Then there are counties that the Democrats once won, as recently as Kerry in 2004, or Obama in 2008, or Gore in 2000, counties predominantly in western Pennsylvania: Beaver, Washington, Cambria, Fayette, Greene, Mercer, Lawrence Counties. Closely related are the large counties in Western Pennsylvania near Pittsburgh that have supported Republican candidates for a longer time: Butler and Westmoreland County for example. Among these counties only Butler experienced any substantial growth in recent years although the population of Washington County slightly increased. The rest are some of the fastest declining counties in terms of population in Pennsylvania.
Another category are the fast-growing counties on the Maryland border, which are closer to Washington D.C. in some cases than they are to Philadelphia: York, Adams, Franklin, Cumberland, Lebanon Counties. Finally there are the many remaining, mostly small counties with population under 100,000 that are, for the most part, declining in population, that have long been Republican strongholds, and continued to support the Republican candidate in the recent election.
Luzerne, Northampton, and Erie Counties
These three counties were successful Democratic counties for many elections but all made sudden and decisive swing towards Donald Trump in the 2016 election. Two of these counties, Luzerne and Northampton, are in northeastern Pennsylvania, near Monroe and Lackawanna Counties which, as I discussed previously, were won by Clinton but with much reduced margins. The pattern continued in these two counties but the reversal in margin resulted in Trump victories in these counties.
The importance of these three counties in the election can hardly be understated: they are respectively, the 12th, 13th, and 14th largest counties in Pennsylvania with a combined population of almost 900,000 residents, and a total of 607,000 registered voters, of whom 406,000 actually voted in the 2016 election. In 2012 the three counties together gave Obama 199,949 (53.49%) votes of the 373,793 votes cast. Clinton, on the other hand received 176,835 (43.56%). Trump received 210,493 (51.85%) to Romney’s 165,901 (44.38%) votes in 2012. These three counties alone saw a swing of over 67,000 votes, greater than Trump’s statewide margin of 44,292 votes. Together with neighboring Monroe and Lackawanna County which, although winning counties for Clinton saw a much reduced margin, these counties effectively decided the election. Clinton’s inability to get the traditional Democratic support of these counties was the absolute deciding factor in the 2016 election. Regardless of the continuing decline in the Democratic vote in western Pennsylvania, or the growing population of conservative counties like York, Franklin, or Cumberland, had Clinton managed to hold onto the traditional margins in the cluster of neighboring counties north of Philadelphia: Berks, Lehigh, Northampton, Monroe, Lackawanna, and Luzerne Counties being the most prominent, The election in Pennsylvania would have been settled early. Similar counties in Michigan and Wisconsin resulted in the same pattern: surprising narrow losses by Clinton to Trump of states that had proven reliably Democratic for decades.
Why should these counties in particular have transformed so rapidly from reliably Democratic to Trump-friendly counties? Interestingly, all of them have similar demographic patterns: Every single one of these counties has a lower fraction of college educated residents than the statewide average of 28.6% (Berks-23.2%, Lehigh-28.5%, Northampton-27.2%, Monroe-23.0%, Lackawanna-25.9%, Luzerne-21.4%). Erie (26.1%), in the far northwestern corner has similar statistics as these contiguous counties near New Jersey. The percentage of Non-White residents correlates very well with the support for Clinton. Monroe (32.4%) and Lehigh (33.1%) Counties, with over 30% Non-White populations, were both counties Clinton won with a plurality of the vote (48.45% and 49.97% respectively). Berks (26.6%) and Northampton (22.1%) Counties, with Non-White populations between 20 and 30%, were losses for Clinton but her total was in the 4o-50 percent range (42.61% and 45.84% respectively), while Luzerne County (15.9%) gave Clinton 38.60% of the vote, the biggest absolute drop (from 51.68% in 2012, a drop of 13.48 points) of any county in Pennsylvania. Lackawanna County (13.0% Non-White) was the only bright spot for Clinton, where she received 49.79% of the vote, but her vote here was down over 13 points from Obama’s 63.09% in 2012, a drop almost as big as the one in neighboring Luzerne County. If anything, one would expect the Democratic party to continue to drop votes in Lackawanna County in future elections. Incidentally Erie County is also a slight anomaly, as the Non-White population in that county is only 15.0% but Clinton captured 46.44% of the vote in the 2016 election. However, like Lackawannna and Luzerne, this was down from 57.36% of the vote in 2012, 10.92 points lower than Obama’s share in 2012.
So, these counties, all with CNW numbers lower than most of the big Democratic counties, all within 10-12 points of the statewide CNW, are the places where the election was lost for the Democratic party. Lehigh (CNW=61.6) and Monroe (CNW=55.4) both had CNW higher than the statewide average and were winning counties for the Democratic party, albeit with reduced margins. Berks (CNW=49.8), Northampton (48.0) and Erie (41.1) all had CNW numbers lower than the statewide average but by less than 10 points; these were all counties won by Obama but lost by Clinton, although her vote share was in the 40s in all three cases (42.18%, 45.84%, and 46.44% respectively; although Obama lost Berks in 2012, he won the county in 2008). Lackawanna (CNW=38.9) and Luzerne (37.3) were both counties in which the Democratic party lost a large share of voters; in the case of Luzerne, Clinton’s support dropped below 40% (38.60%), while in Lackawanna, Clinton managed to eke out a victory (49.79%).
The White Surge
Collectively the sixteen counties already discussed make up 66.58% of the vote and 65.19% of the population. Critically, although Clinton maintained a lead of 690,521 votes over Trump, Obama had a lead in the same counties of 747,772 votes over Romney in 2012. This difference of 57,251 votes in the margin of the sixteen counties discussed is the critical difference between Clinton winning and losing the election. Clinton at this point had 31,877 more votes than Obama in 2012. However, Trump had increased his vote total from Romney’s in these same counties by 89,128 votes. This critical surge of voters, mostly in the counties to the north of Philadelphia was the key to the victory of Donald Trump in Pennsylvania. The number of voters increased in all these counties from 2012: In Berks County turnout was up 14,000, Trump was up 12,000 votes from Romney; Luzerne County up 11,000, Trump up 20,000 votes!; Northampton up 14,000, Trump up 10,000 votes; Lackawanna up 6,000, Trump up 13,000 votes; Monroe up 8,000, Trump up 7000 votes, and so on.
The remaining counties in Pennsylvania comprised approximately one third of the electorate. As it happened, voter turnout in the remaining third of Pennsylvania was 8.8% higher than in 2012, compared with a 6.6% rise in turnout in the sixteen counties already discussed. As I have pointed out above, the swing in the vote of the counties to the north of Philadelphia was the decisive factor in Clinton losing the election. However, it is also true that the surge in voting in the remaining third of the electorate was the decisive factor in cutting Obama’s margin of 310,000 votes. Clinton received only 617,189 votes in the remaining 51 counties (29.94% of the vote!), while Trump garnered 1,352,002 votes in these same counties (65.60%). Obama, although he also fared poorly in these counties, managed to get 712,899 votes, 96,000 more than Clinton, while Romney managed 1,150,831 votes, 200,000 fewer votes than Trump. In other words over 100,000 voters that did not vote in 2012 turned out in these counties in 2016 to cast a ballot for Donald Trump, while another 50,000 cast ballots for other candidates besides Clinton or Trump.
Of the remaining 51 counties, only three (York, Cumberland, and Pike, discussed below) have a non-White population greater than 10-12%. The remaining counties virtually all have a population that is over 90% White. Thus, the surge in voters in this remaining set of counties by definition must be a surge of White voters, much as Sean Trende surmised had gone missing in the 2012 election. He suggested that these voters come from counties in which Ross Perot did well in 1992.
It is apparent in every few elections, when voter turnout suddenly surges, that there are many people who only occasionally vote in Presidential elections. In 1992, for example, voter turnout in Pennsylvania surged from 4.54 million voters in 1988 to 4.96 million voters. Ross Perot, an Independent candidate with a protectionist economic message, garnered 902,000 votes in Pennsylvania in the 1992 election, mostly at the expense of George Bush, whose vote total dropped by 500,000 from 1988. Bill Clinton, the incidental beneficiary of this drop, earned 2.24 million votes, a mere 45,000 votes more than Mike Dukakis’s 2.19 million votes, and Dukakis lost by over 100,000 votes in 1988. The 400,000 new voters clearly went for Ross Perot (if you add 500,000 + 400,000 you get 900,000: simple!).
In 1996 Perot ran again: Voter turnout returned to 1988 levels (4.51 million votes cast). Clinton got virtually the same number of votes as he did in 1992 (2.22 million votes) and the Republican candidate Bob Dole got 1.80 million votes, only 10,000 votes more than George Bush in 1992. Perot garnered 431,000 votes, more than 470,000 fewer votes than his 1992 total. Most of the 400,000+ voters from 1992 had once again disappeared and Perot’s vote total was cut in half.
Interestingly, in the 2000 election George W. Bush earned almost as many votes (2.28 million) as his father did in 1988. However, Al Gore increased his vote total to 2.49 million, 250,000 more votes than Clinton and more than any Democrat had earned since Lyndon Johnson’s crushing defeat of Goldwater in 1964. By 2004 both parties had improved their ‘get out the vote’ operations and voting levels reached heights unseen in the history of Pennsylvania, where they have essentially remained to the present day.
The main challenge is to determine if the 400,000 or so difference in the voters who turned out in 2016 but were absent in 2012 is likely to be repeated. History indicates that some of the vote will disappear as enthusiasm wanes (see Perot 1996). Additionally, the counties in which the surge was most prominent, specifically the remaining 51 counties and the half dozen or so counties in northeastern Pennsylvania I have already discussed, are some of the counties with biggest population declines in recent years. I suspect that voting levels will drop off in 2020 and that the biggest drop will occur in the 3 or 4 counties just discussed (i.e Luzerne County etc.) and in the majority of the remaining 51 counties I will now examine.
The single largest batch of remaining votes is to found in the counties surrounding Pittsburgh. The Democratic party has been losing votes for a long time in western Pennsylvania. Where Al Gore won all but one (Crawford being the exception) of the counties bordering Ohio and West Virginia including Erie, Mercer, Lawrence, Beaver, Washington, Greene, Fayette, as well as winning Cambria County to the east of Pittsburgh, only Pittsburgh in Allegheny County has remained a Democratic stronghold. Even Bill Clinton, Michael Dukakis, and Walter Mondale prevailed in these counties, as well as in Westmoreland County and one or two more counties in the western part of the state. Ironically, Democrats before Bill Clinton managed to hold only Philadelphia in eastern Pennsylvania: every other county in the east and central part of Pennsylvania went to the Republicans in the elections prior to the 1990s.
However, the situation has reversed and now Southeastern Pennsylvania is the strong area of the state for the Democrats, while the western part of the state, save Pittsburgh, is controlled by the Republicans. The numbers get worse with each election. The only drawback for Republicans is that, with the exception of Butler County to the north of Pittsburgh, none of these counties is growing in population and thus represent a limited supply of potential future voters. Essentially the Republicans can only cannibalize the declining pool of Democratic voters and hope the Republican voters are long-lived, which is highly unlikely in towns whose economies long been based on coal, steel, and heavy industry, and whose citizens are some of the unhealthiest and oldest in the nation.
The single largest county by population in Western Pennsylvania after Allegheny is Westmoreland County, which borders Allegheny to the east. This county is the 11th largest county in the state, with a population of 357,956 in 2015, down from 365,169 in 2010 and 369,993 in the 2000 US Census. Westmoreland and neighboring Cambria County are the counties with the largest population decline in the recent five year period recorded by the US Census Bureau. Both counties lost over 7,000 people in a mere five year period. Despite this fact, Westmoreland County had an increase in registered voters from 2012 of 8,000 (246,020 versus 238,006) a number lower than the 249,140 registered voters of 2008 and the high of 256,365 registered voters in 2004.
More remarkable still, the number of voters to cast ballots in Westmoreland County reached an all-time high of 183,492, almost 5,000 voters more than the previous largest number of 178,696 in 2004. This turnout represented 74.6% of all registered voters, one of the highest turnouts in Pennsylvania, and an impressive 51.26% of the entire population of the county. For comparison, Philadelphia’s 709,618 voters cast represents 45.27% of the population of the city, although high-voting Chester County saw 52.91% of it’s population (272,998 of 515,939) vote in the 2016 election (77.0% of registered voters). Those who turned out in Westmoreland County overwhelmingly supported Trump, giving him a record 116,427 votes (63.50% of the vote) compared to Clinton’s 59,506 votes (32.52%), the lowest vote total for a Democratic presidential candidate since George McGovern in 1972. This in a county in which Walter Mondale collected 79,906 votes to defeat Ronald Reagan (71,377 votes). If it could get any worse I would be surprised.
Except Clinton fared even worse in neighboring Cambria County, earning a mere 29.67% of the vote in a county Barack Obama narrowly won in 2008 over John McCain! The numbers are similar in virtually every one of the counties surrounding Pittsburgh. Beaver: 38.5% for Clinton, 47.6% for Obama in 2008, 52.9% for Gore in 2000; Washington: 35.5% for Clinton, 46.8% for Obama in 2008, 53.2% for Gore in 2000; Lawrence: 34.1% for Clinton, 46.5% for Obama in 2008, 52.0% for Gore in 2000; Mercer: 35.5% for Clinton, 48.8% for Obama in 2008, 48.9% for Gore in 2000; Fayette 33.2% for Clinton, 48.9% for Obama in 2008, 56.8% for Gore in 2000; Greene: a stunningly poor 28.2% for Clinton, 48.6% for Obama in 2008, 53.0% for Gore in 2000.
These counties are gone and they are not coming back into the Democratic fold for a long time. These counties are following trends seen in other parts of the country: Kentucky, West Virginia, Ohio, etc. where predominantly White, less well-educated, aging counties, increasingly vote Republican. While I would love to think there is a magic formula that will bring back these voters into the Democratic fold, I fear that years of Fox propaganda about ‘urban takers who abuse government entitlement programs and generally get a free ride on the backs of the working class’ has fed into a simmering resentment of the urban coalition of highly-educated liberal ‘elite’, and non-White voters. In the zero-sum world that they believe exists there must be must be winners and losers and the battle lines have unfortunately been drawn. Barring a (not at all out of the realm of possibility these days) cataclysmic event, the long term direction of the country will not favor the voters in these counties, particularly now that the level of hostility has been cranked up to eleven.
For purposes of this discussion, I have grouped 12 counties in Western Pennsylvania into this category: Mercer, Lawrence, Beaver, Washington, Greene, Fayette, Westmoreland, Butler, Armstrong, Indiana, Cambria, and Blair Counties. Collectively these counties had a population of 1,760,376 inhabitants as of 2015, about 14% of the Pennsylvania total. Not one county voted for Clinton; the highest total she managed was 38.5% in Beaver County, while the lowest was a paltry 22.5% in Armstrong County.
The percentage of college graduates in these counties range from a low of 13.9% in Armstrong to a high of 32.3% in Butler County. The CNW in these counties ranges from a low of 17.4 in Armstrong County to a high of 37.2 in Butler County. The only county among the twelve with a Non-White population greater than 10% is Beaver County which is also the county in which Clinton had her best performance. The highest White population is in Armstrong County, where only 2.8% of the population is Non-White. Coincidence?
York, Cumberland, Lebanon, Franklin, and Adams County
These counties represent one of the few bright spots demographically for the Republican party. All five of these counties are growing and all five are heavily Republican in their voting preference. Clinton failed to reach the 40% threshold in any of these counties, which have a combined population of 1,082,205 people, 8.5% of the total population of Pennsylvania. These five counties grew by a net 27,234 residents in the five years from 2010-2015, a 2.6% increase in a state that grew by 0.8% in the same period.
York County is the 8th largest county in Pennsylvania, with 442,867 inhabitants in 2015. Lancaster County is to the east, Dauphin County to the north, and Maryland to the south. Along with Franklin, Adams,and Cumberland Counties bordering York to the west and nearby Lebanon County between Dauphin and Lancaster County, these counties make up the last area in Pennsylvania with a positive growth trend. York added almost 8,000 net residents in the five years from 2010, while Cumberland added almost 11,000 net new residents, Franklin County’s population increased by 4,000 residents, Adams County by a little under 1,000, and Lebanon County by 3,500. This is an area that has been and continued to be fertile ground for Republicans and in theory provides the sole area of potential population growth for the Republicans (as opposed to pulling Democrats from counties with declining populations).
Clinton failed to reach 40% of the vote in any of these counties, although the 38.13% of the vote she earned in Cumberland County was nearly as high as the percentage she earned in Luzerne County (38.60%), a county won by Obama in 2012. Cumberland is one of the six counties in the state with a CNW>40 (45.7). After Berks and Northampton, this is the highest CNW county not to vote for Clinton and, adjacent to Dauphin and fast growing, seems to have potential to become more favorable to the Democratic party in the near future. In fact, Clinton’s vote total in Cumberland was higher than Obama’s in 2012, the only one of these five counties in which the Democratic vote total increased from the previous election.
The remaining counties of Pennsylvania
The remaining 34 counties, although making up half of all the counties in the state and the majority of the actual area, comprise about 1.6 million people, roughly the size of Philadelphia. about 12.5% of the total population. There is one county in this group that has a CNW>40, Pike County north of Monroe, on the New Jersey border and the only county in Pennsylvania considered as part of the New York Metropolitan Area. Every county in this group voted for Trump and no county but Pike (at 35.2%) gave Clinton more than 35% of the vote. Almost every county is declining in population. Although turnout was up in these counties compared to 2012, the likelihood of the Republicans squeezing many more votes out of these counties is small.
Not one county in this group has a population larger than 150,000 residents, and almost every single one is losing population. Schuykill and Lycoming counties are the only ones over 100,000!! Schuykill is a county which is between many of the clusters I have created for these articles but which fits truly into none of them. It is in the middle between Dauphin, Lebanon, Luzerne and Berks counties. Perhaps it will fit better into another category but for now it is the largest remaining county in the ‘leftover’ group. Clinton received 26.5% of the vote, a not atypical result in the majority of these counties. The main function electorally of these counties seems to be to counteract the vote of Philadelphia. The vote in Allegheny County in western Pennsylvania is neutralized by the vote of the surrounding counties, and the vote of the counties like York, Lancaster, Luzerne, and other counties in the far outer ring of Philadelphia neutralize the vote of the counties in the inner ring, such as Delaware and Montgomery Counties.
This seems to be the situation in which Trump managed to pull off a surprise victory: by increasing turnout in the ‘neutralizing’ counties and reversing the vote in historically Democratic counties like Luzerne and Lackawanna, combined with a large segment of the voting public casting ballots for third party candidates, Trump managed to eke out a small victory in Pennsylvania, surprising the pundits and confounding the polls showing Clinton to be ahead in the final days of the campaign.
Toward a Winning Strategy
To reiterate: ten counties had a CNW>50 (Pennsylvania statewide average 50.7) and Clinton won all ten counties. Six additional counties had a CNW between 40 and 50 in 2015: Clinton lost all those counties, but three of these, Berks (CNW=49.8, 42.61% of vote), Northampton (CNW=48.0, 45.84% of the vote), and Erie (CNW=41.1, 46.44% of the vote) were counties in which Clinton got more than 40% of the vote. In the other three counties, Lancaster (CNW=42.6, 37.21% of the vote), Cumberland County (near Dauphin County, CNW=45.7, 38.13% of the vote), and Pike County (neighboring Monroe County, CNW=42.7, 35.22% of the vote), Clinton received between 35 and 40% of the vote. Lackawanna County (CNW=38.9) was the highest CNW county under 40 and Clinton won that county with 49.79% of the vote. All the remaining 50 counties had a CNW below Lackawanna County (i.e under 38.9) and Clinton got less than 40% of the vote in every single county!
The ten counties with CNW>50 had a total population of 6,284,710 residents, which is 49.1% of the population of Pennsylvania and includes 7 of the state’s ten largest counties. The net population increase in these counties from 2010 to 2015 was 111,189 net new residents; thus the remaining 57 counties collectively lost population in the same five year period. The six counties with CNW numbers between 40 and 50 have about 1.8 million people between them, about 14% of the total population and include two of the top ten largest counties. Four of these six counties gained population recently and the net population gain of these six counties between 2010 and 2015 was about 31,000. A further three counties, Lackawanna (CNW=38.9), York (CNW=38.3) and Luzerne County (CNW=37.3) have CNW numbers greater than 37; the Democrats traditionally have won two of the three, but lost Luzerne this election, and are likely to lose Lackawanna in future if current trends continue. York added 8,000 people while Lackawanna and Luzerne collectively lost about 5,000 people.
York, Luzerne, and Lackawanna are included among the counties that at at least theoretically battleground territory based on past results and based on the fact that they are the highest CNW counties under CNW=40. Taken together they have a population of 973,000, 7.6% of the Pennsylvania total. Thus, the 19 counties with a CNW>37.3 collectively contain 9.1 million people, 71.0% of the total population, and include 16 of the 17 largest counties in the state (Westmoreland County at #11 being the sole exception). Fourteen of the nineteen are growing counties; Clinton lost three of the five counties that are declining in population, and won nine of the fourteen counties that are growing.
Only one of the counties not discussed in this section thus far (which, by definition, means it has a CNW under 37.3) has a population whose education levels exceed that of Pennsylvania (28.6% bachelor’s degrees) as a whole. The population of Butler County, immediately to the north of Pittsburgh has a relatively high proportion of bachelor’s degrees at 32.3% . It is, however, 95.1% White, meaning the Non-White Population is a mere 4.9% of the total. Notice that the CNW of Butler County is 37.2, just below Luzerne County as the next county in line ranked by CNW. Trump received 65.71% of the vote in Butler County to Clinton’s 29.15%, so it seems that, in terms of discussing potential battleground counties, Butler seems to mark the end of any competitive counties in the state. Every other county in Pennsylvania has a population below the state average of 28.6% bachelor’s degrees.
One or two additional counties have a Non-White population greater than that of Lackawanna County (i.e more than 13.0%: Lebanon County, at 16.5% is the only sizeable county (137,o67 residents in 2015) with a reasonably large minority population. One other county, Union County (adjacent to Centre County) has only 45,000 people but has a 15.5% minority population: However, further investigation revealed the presence of Lewisburg Federal Penitentiary. Kelly Township, the actual location of the prison, has a very large minority population, many of whom are very likely to be residents of the prison and thus unable to vote. Similarly, tiny Forest County has a 26.0% minority population and is also the home of a large prison. Otherwise, there are no counties remaining in Pennsylvania with a minority population of more than 12% of the total.
Thus, in my view, the most useful strategy for the Democratic party would be to focus on the roughly twenty counties with a CNW of 37 or higher. Nineteen of these counties are contiguous and form a rough arc from the northeast to the southwest of Philadelphia; the other is Allegheny County (one would like to think that the increasing liberal voters of Pittsburgh might eventually increase and spread into neighboring counties, but I am not holding my breath). The strategy should aim to consolidate and expand the vote in the larger growing and Democratic core counties like Philadelphia, Allegheny, Delaware, and Montgomery; strengthen the margins in counties like Chester and Bucks County, which will increasingly come to resemble Montgomery and Delaware Counties; focus attention on counties where Clinton reached 40-50% of the vote and have strong demographic upsides like Berks County, Dauphin, Lehigh, Monroe, Northampton, and perhaps even Pike County; And work very hard on the growing area southwest of Philadelphia, particularly Lancaster County, but also York, Cumberland, and perhaps even Lebanon County. Only after all these strategies have been employed should the party chase the White voters who left for Trump in 2016: Luzerne, Erie, Lackawanna are all still winnable if a serious Democratic candidate can make the case that he or she will create jobs and help those left behind with proposals that will help everyone including the working poor in urban areas, who are often more likely to not be White. A direct appeal to White working class men alone will fail in the long run because it will alienate other richer sources of future voters. The appeal needs to be more like Bernie Sanders’ message of concern for all workers and anger at the wealthy for abusing and manipulating the system for personal gain.
2020: A Hypothetical Election
There are many ways to think about how the 2020 election might unfold. Perhaps Trump will be successful enough that, as an incumbent, voters will not want to change horses in midstream. This seems fanciful but is a hypothetical possibility. In that event, the Democratic party will hang on until 2024 to try to take advantage of what, by then, will be a radically more favorable demographic environment. Or, a more likely possibility, the aggrieved Democratic voters who have suffered four years of terrible government, will focus like a laser beam on getting Trump out and will be energized. This is also a possibility, but it would require a decent candidate that the party can unite around and none seems to be on the horizon. The most likely scenario is a blend of the two; some voters will remain resigned to the status quo, others will be energetic in their opposition.
The one thing we can do is try to make a prediction based on the likely demographic changes between now and 2020. How will the change in the population affect the makeup of the electorate? One thing is certain: The counties favorable to Democratic candidates (especially those with a CNW>50), are growing. Among those that could be potential targets for the Democratic party (CNW>37) Berks, Northampton, Lancaster, York, Pike, Cumberland are all growing, while Lackawanna, Luzerne, and Erie are shrinking. Very few of the remaining counties are growing and the overall trajectory for the remaining counties with CNW<37 is population decline. Thus it seems unlikely that the high turnouts reached in 2016 can be matched in four years time in counties like Westmoreland: perhaps the percentage turnout might be as high, but the actual number of likely voters seems to have reached saturation point in 2016.
Therefore, my first assumption is that the CNW<50 counties will have a smaller share of the electorate. Let’s assume the turnout rate is similar in 2020 to the turnout rate in 2016. In 2016 the 12 counties with CNW>48.0, which includes Berks and Northampton County as well as the ten ‘Clinton’ counties, had a total of 7 million residents, about 55% of the state’s population. Of these, 3.5 million people, almost exactly 50% of the population, cast a ballot in 2016. This group of counties had a net population increase from 2010 to 2015 of about 120,000 people, meaning that the remaining 55 counties in Pennsylvania collectively lost a net 20,000 people. There is little reason to think the trend will do anything but accelerate in the five years to 2020. Therefore in 2020, the 12 ‘Democratic’ leaning counties should have at least an additional 100,000-150,000 more residents. If 50% of them vote, one would expect at a minimum 3.55-3.60 million voters to cast ballots; in other words, there might be an additional 50-100,000 voters, assuming the rate of voter turnout stayed exactly the same
Similarly, assuming the remaining counties continue on their slow downward spiral, the population will drop slightly from 5.8 million to perhaps 5.75 million. In 2016, these counties collectively cast 43.35% of the votes, although the population represents 45.32% of the total. Perhaps the Census data is behind actual population trends in Pennsylvania, or perhaps rural voters vote at slightly lower rates, but for some reason the voters in 2016 represented only 46% of the population. Despite the ‘surge’ of voters in 2016, this general pattern is no different from previous elections, where the total voting population slightly lags that of the larger counties as a percentage of the population. Thus I will go with the 46% number and use that to estimate that the number of voters in 2020 should be about 2.60-2.65 million, slightly below the 2016 total of 2.67 million. Thus, if the rate of voter turnout remains constant, the 55 counties that are less friendly to Democrats will have a marginally smaller turnout.
The next thing we need to look at is the likely voting habits of each group of voters. In 2016, the twelve ‘high CNW’ counties, which from now on I will call Democratic counties for simplicity, gave 59% of their vote to Clinton and 37% of the vote to Trump. On the other hand the 55 ‘low CNW’ counties, the ‘Republican’ counties, gave almost 63% of their support to Trump and a little under 33% to Clinton. If nothing changed except for population changes as described above the results would look something like the following: Assume total voters at 6.2 million (3.6m+2.6m). If voting percentages for each group were to remain constant the Democratic candidate would get 2.98 million votes while the Republican would get 2.97 million votes. In other words, holding everything else constant and changing only the population based on census projections, the Democrats would likely prevail by a slim margin in the 2020 election in Pennsylvania.
The problem is, of course, that nothing stays quite constant. As this series of articles should make abundantly clear, different groups are in the process of evolving in terms of their voting patterns, the makeup of each of the two major groups is likely to undergo an evolution as well, even if it is a small one. To take one example, In Montgomery County between 2010 and 2015, the population increased by a net 19,000 residents. The total number of White (Not-Hispanic) residents declined from 633,000 to 628,000, while the number of self-identified Asians increased from 52,000 to 62,000, the number of Hispanics increased from 34,000 to 40,000 and the number of Blacks increased from 68,000 to 74,000; in total the non-White groups increased by 24,000 while the White population declined by 5,000 people. Although the data is missing here, the makeup of the White population is likely to have undergone a transformation as well, as younger, more educated people replace older people with lower rates of college degrees.
This is part of the reason that Montgomery continues to move in the direction of the Democratic party, and this pattern is true not just of fast-growing counties like Montgomery, but also of ‘Republican’ counties like Westmoreland: there the population dropped by 7,000 residents but by 9,000 White residents with balance being an increase in the non-White population. Thus, it would be naive to assume the voting patterns will stay the same in 2020. What is more likely to happen is that the vote in ‘Democratic’ counties will trend more in favor of the Democrats. In the ‘Republican’ counties’ the difficulty is that, as the population ages, the vote seems increasingly to trend towards the Republican candidate, particularly in western Pennsylvania. However, this is somewhat offset by the simple fact that the population is declining through attrition, and that the rate of White decline is greater still, as I illustrated above. Thus, it seems likely that even if the newcomers do not vote, the best outcome in these ‘Republican’ counties is that the Democratic candidate loses votes to the Republican but that the Republican totals fail to increase much owing to population decline. As I said previously, it seems like fighting for a bigger slice of a shrinking pie.
Add to this the fact that I have deliberately excluded from the ‘Democratic’ column counties that are slowly trending towards the Democratic party, including the largest single county in the ‘Republican’ group, Lancaster County, and it is hard to imagine that the Republicans will be able to even match their highs of 2016. I remind the reader that, since 2004, the Democratic ticket has increased its vote totals from 74,328 to 91,093 in Lancaster County, while the Republicans have declined from 145,591 votes in 2004 to 137,914 votes in 2016. I see this trend continuing as Lancaster grows and becomes ever more diverse.
A more accurate prediction then would factor in the above changes. Even a small tweak in the result makes a big difference. For example, let’s argue that the Democratic candidate wins 60% of the vote in the ‘Democratic’ counties while the Republican drops to 36%. Then let us keep the ‘Republican’ county result constant, for the reasons I just gave. This will obviously favor the Democratic candidate, who now stands to garner 3.02 million votes to the Republican’s 2.93 million, an 80,000 vote victory for the Democratic candidate, keeping everything else constant.
I could go on with these hypothetical examples, but the main point is that if the Republicans hope to win the state of Pennsylvania, they will very likely not be able to rely on increasing turnout to a level as high or higher than that of 2016. In fact, the opposite is likely to happen: some of the enthusiasm will have diminished among voters who turned out for Trump in 2016, much like the case of Ross Perot in 1996. In all likelihood, ‘Republican’ counties will have a lower turnout, even if there is a continued shift away from the Democratic party. On the other hand, population growth alone in the Democratic counties, coupled with resentment and anger at the current President, will likely increase or at least stabilize turnout in ‘Democratic’ counties.
I know I was wrong in 2016, as I failed to take the ‘missing white voter’ hypothesis seriously enough, but the data suggests strongly that a repeat performance would be very difficult indeed to pull off in 2020. I predict that Pennsylvania will revert to the Democratic column in 2020.