General Overview
Pennsylvania is a Mid-Atlantic state with a population of 12.96 million. 35% of Pennsylvanians hold a bachelor’s degree or higher (about even with the 36% nationwide statistic), while 33% hold only a high school diploma. The state is 81% white, and 74% white excluding Hispanic populations (indicating a significantly higher white population than the 75% and 58% nationwide numbers respectively). It is part of America’s Rust Belt, a term for the Midwestern to Mid-Atlantic region known for being a stronghold of industry and manufacturing that has been in decline since the 1950s.
Considered one of the country’s most pivotal swing states in presidential contests, Pennsylvania often hosts one of the closest races (and as a result, is one of the most difficult states to predict), and frequently votes for the winner of nationwide elections. Its 19 electoral votes can often be essential to a winning campaign, making it a strong focus for candidates in an increasingly calcified political climate. Down-ballot, Pennsylvania currently has a Democratic governor, a Republican secretary of state, two Democratic senators, a split cohort of congressional representatives, and a split state legislature. Several of these offices see frequent changes in party control.
Forecasted Outcomes vs. Actual Outcomes
My hybrid fundamentals-polling model predicted incorrectly that Kamala Harris would win Pennsylvania with a 51.17% share of the two-party vote, but this was well within the margin of error. For comparison, Biden won the state in 2020 with 3,458,229 votes to Trump’s 3,377,674, or 50.5%.
The model’s prediction that she would win with a greater margin than Biden did turned out to be quite off-base. In actuality, 3,421,247 votes were cast for Harris in 2024, while 3,542,701 were cast for Trump, giving the latter the win with 50.8% of the two-party vote. The former president flipped four Pennsylvania counties from blue in 2020 to red in 2024: the notoriously competitive Erie County was among them, along with the counties of Monroe, Northampton, and Bucks.
My Pennsylvania forecast was wrong along with those of experts on the fundamentals and polling sides. For one, FiveThirtyEight predicted Harris would win the state with 47.9% to Trump’s 47.7% of the total vote – interestingly, this forecast suggests 95.6% of voters would cast ballots for one of the two major parties, while the actual result had Harris at 48.7% to Trump’s 50.4% of total votes, with a higher 99.1% of voters opting for one of the two parties. It appears that a central mistake of FiveThirtyEight’s forecast in the state could have been overestimating third-party votes, and it seems that these unexpected non-third-party voters swung more towards Trump than Harris. This could be an explanation for the overperformance of polling we have seen from Trump in every election– voters leaning towards a third party may make late decisions to vote for Trump, or simply are reluctant to disclose their eventual plans.
2020 Map: (https://www.politicspa.com/fun-with-maps-election-night-2024/139796/)
2024 Map: (https://www.politico.com/2024-election/results/pennsylvania/)
Campaign Narrative & Analysis
As early as May 2024, Pennsylvania voters showed signs of turning away from the Democratic Party, not only because of Joe Biden’s lackluster showing as a returning candidate, but because of the issues themselves.
This poll from Inquirer/NYT/Siena (https://www.inquirer.com/politics/election/trump-biden-pennsylvania-issues-poll-20240513.html) showed that Biden trailed Trump not only generally, but when Pennsylvanians were surveyed on individual issues. The only issue where Biden outperformed Trump was that of abortion – Trump was more trusted by Pennsylvanians to handle the economy, foreign policy issues, and most other pressing problems.
Biden also lost significant ground among key Democratic base demographics. Among young voters, his support dropped from 62% in October 2020 to 47% in May 2024, while it fell from 71% to 57% of non-white voters over the same time frame. The issues poll conducted is an important one in hindsight, because it suggests that criticisms of Biden went beyond his age and apparently declining mental ability, but may have revealed more substantial critiques of Democratic policy in these areas – leanings that may not have actually vanished once a new candidate entered the picture.
Poor perceptions of the economy contributed to these statistics, but some scholars think it is still possible to win in such conditions. In her book The Message Matters, Lynn Vavreck suggests that economic conditions can easily sort candidates into the type of campaign they need to run. If you are the incumbent in a “good” economy or the challenger in a “bad” one, the economic conditions favor you, and you should run what she terms a “clarifying campaign,” focused on economic issues that are most central to voters’ daily lives. If you are the incumbent in a “bad” economy (or the challenger in a good one), you should run an “insurgent” campaign, focusing on a non-economic policy issue that resonates significantly with voters. The Democratic candidate, whether Biden or Harris, would need to run an insurgent campaign to be successful, but even so, odds are typically stacked against the insurgent candidate.
Once the president dropped out after national pressure following a particularly weak debate performance, Pennsylvanians, along with the rest of the country, briefly showed more support for the Democrats. At the conclusion of the Democratic National Convention, where Harris was officially named the nominee, she was up in FiveThirtyEight’s polls by a full two percentage points (as compared to the slight 0.2% advantage she held by the end). In order to keep up the momentum, however, Vavreck would have theorized that Harris needed to continue running an insurgent campaign. Abortion was ranked as a top issue among Pennsylvanians, suggesting potential to run on popular Democratic policy in that area (https://www.spotlightpa.org/news/2024/09/abortion-presidential-election-poll-2024-pennsylvania-kamala-harris-donald-trump/), but the issue would need to be sufficient virtually alone to turn out enough base and swing voters, which may be a tall order.
A key part of the campaign in Pennsylvania was the weighting of Harris’ running mate selection. Pennsylvania Governor Josh Shapiro was among the leading candidates for the position, considered for his strong track record and popularity with Pennsylvanians. However, the selection of Minnesota Governor Tim Walz over Shapiro caused some to wonder whether Harris would have won Pennsylvania had she selected its governor as her running mate.
Looking back on 2022, Shapiro won the governor’s race with 56% of the vote, or 3,031,137 votes. By contrast, 3,421,247 votes were cast for Harris in 2024, while 3,542,701 were cast for Trump. Harris received more votes in 2024 than Shapiro did in 2022, but still lost her race while Shapiro won his. This suggests that the difference in Republican turnout was key; in other words, a large group of Trump-only voters that did not care to vote for the Republican gubernatorial candidate may have turned out in 2024 regardless. Even if every Shapiro voter had voted for Harris, it may not have made the difference. Therefore, it is hard to conclude that selecting him would have boosted her margins enough to flip the state.
As the general election loomed closer, both candidates made significant ground game, air war, social media, and texting efforts in Pennsylvania, especially in perceived battleground areas.
Erie County, located at the northwest edge of Pennsylvania, was perceived as a “battleground within the battleground.” The county almost always votes for the winner of presidential elections; In 2020, Biden won the county with just 49.81% of the county’s support: he received 68,286 votes, compared to 66,869 for Trump. Some consider Erie a “microcosm of the nation,” although it is “whiter… poorer… and less educated” than the rest of Pennsylvania. To turn out and persuade this most variable voter base, both campaigns focused many of their efforts in Erie County. (https://www.voanews.com/a/pennsylvania-s-erie-county-is-pivotal-in-presidential-race-/7844886.html) (https://www.npr.org/2024/10/05/nx-s1-5129676/erie-county-pa-trump-harris-2024)
Erie citizens, as well as Pennsylvanians at large, reported being inundated with campaign messaging from both sides, through text and TV among other platforms. The state held the tightest, most contentious, and most expensive race of any state in the country. By October, Democrats had spent 159 million dollars on ads alone in the state while Republicans spent 121 million. (https://www.npr.org/2024/10/21/g-s1-28936/pennsylvania-election-harris-trump-2024)
Noticing Biden’s decisive victory among mail-in voters in 2020, Erie’s Republican campaign strategists focused on improving mail-in turnout amongst their base.
Despite an aggressive air campaign and significant fundraising efforts, some Harris campaign volunteers thought the Democrats’ ground game in Pennsylvania was lacking. Field offices appear to have been kept in non-ideal conditions, while volunteers complained that the campaign focused on wealthier white voters to the detriment of turnout efforts in racially diverse neighborhoods. Determined volunteers created a secondary, shadow-like campaign operation aimed at reaching these neighborhoods. (https://www.nytimes.com/2024/12/07/us/politics/harris-philadelphia-black-latino-voters.html?smid=nytcore-ios-share&referringSource=articleShare) In the absence of a stronger Harris campaign effort in this area, Pennsylvania Republicans seized the opportunity to try to persuade Black men towards their cause. (https://www.nytimes.com/2024/06/10/us/politics/2024-election-gop-black-men-voters.html) Ground game can be important when reaching voters face-to-face could convince them of a preferred candidate or even just to get to the polls – if ground game efforts are not distributed well, missed opportunities could sink a campaign that already faces an uphill battle.
After the initial momentum from the candidate switch wore off, the race appeared to tighten further, with Trump briefly overtaking Harris around the end of October.
I also created a programmatic text analysis of all Pennsylvania speeches by both candidates recorded on (www.presidency.ucsb.edu) beginning in September. My initial qualitative observations, just reading through both candidates’ speeches in the area, were as follows: Although Trump delivered fewer speeches, Harris delivered much shorter speeches than Trump, covering much less ground topically per speech. At the same time, there’s also quite a bit of somewhat irrelevant ground covered by Trump speeches. Harris speeches focused more on domestic policy while Trump honed in on the economy, an observation in line with Vavreck’s theory.
After looking over the data, I then used Python to scrape, sort, and format the speeches and called the Gemini API in R to employ a large language model to analyze topics of the set of speeches from Vavreck’s provided categories (traits, economic, domestic, foreign, defense). For Harris, the output was decisive, in that domestic policy the central topic of most speeches with candidate traits a far-off second. For Trump, he mostly spoke on economy but also had a good mix with domestic policy.
Although both candidates did appear to run the types of campaigns Vavreck would have wanted them to, the success of Trump’s clarifying campaign depended on negative perceptions of the current economy and buy-in to the Republican platform, while the success of Harris’ insurgent campaign depended on a sufficient insurgent issue, which she may have lacked.
But even if Harris had run a perfect campaign, she may have been outmatched by the fundamentals. Despite both candidates employing these strategies and spending millions to get their strategically aligned messages out to Pennsylvanians, other scholarship tends to suggest that all this campaigning may have a limited effect in certain situations.
In The Minimal Persuasive Effects of Campaign Contact in General Elections: Evidence from 49 Field Experiments, the authors suggest that there are two main scenarios in which campaigns have a significant effect: > “First, when candidates take unusually unpopular positions and campaigns invest unusually heavily in identifying persuadable voters. Second, when campaigns contact voters long before election day and measure effects immediately—although this early persuasion decays.”
Neither campaign particularly fits the first description – very few policies were proposed by either candidate that could be considered outside of their respective parties’ status quo (which some regard as a significant mistake by Harris). The second description suggests that any persuasive effect under normal circumstances would happen long before the election and decay by the election as voters become more inundated with media. However, even if initial effects decay by Election Day, a higher performance in early polling sparked by initial persuasion might have spurred greater momentum. The Trump campaign had this opportunity to convince voters even up to four years ahead of time, while the Harris campaign lacked the chance to do the same earlier than a couple months out from Election Day. Subscribing to this theory, the Trump campaign may have had a momentum-building effect early on, but the Harris campaign may not have.
The authors also mention that being overly inundated with political messaging decreases the effect of a given piece of messaging, suggesting that the overwhelmed Pennsylvanians in Erie County and elsewhere may have become impervious to campaign messaging altogether. These factors all decreased Harris’ ability to run a successful insurgent campaign.
Meanwhile, the Siena polls (https://www.nytimes.com/2024/11/03/us/politics/harris-trump-times-siena-poll.html) still showed the candidates neck and neck in Pennsylvania up until the end. Towards the final stretch of the campaign process, Elon Musk offered a $1 million lottery open to supporters of him and Trump who registered to vote in Pennsylvania. This was challenged but upheld in court, with the defense arguing that winners were “paid spokespeople” (https://apnews.com/article/musk-million-sweepstakes-lottery-pennsylvania-krasner-4f683c48eb7dcc57f183e54ef16e7320). The effect of this could have been to incentivize registration among Trump supporters, and with such close margins in the state.
Down ballot, three-time incumbent Senator Bob Casey met a similar fate as Harris – in the closest Senate election nationwide, he was narrowly defeated by a Republican challenger. At the same time, the Democrats in Pennsylvania’s state senate succeeded at retaining their majority. One article (https://www.wesa.fm/politics-government/2024-12-06/pa-democrats-down-ballot) looks into PA state legislature Democratic successes and their differences from federal campaigns. State-legislature candidates talked about diverging on messaging and taking a stronger economic-populist stance than the federal candidates did. However, the article does mention that incumbent Senator Casey also ran on an economic populist message and still narrowly lost. If we take the state legislators’ central argument that the messaging was not economic enough, we see that it goes against Vavreck’s notion of an insurgent, non-economic campaign as necessary for the incumbent party in a bad economy.
One might wonder who, if anyone, has successfully run an insurgent campaign. Vavreck’s list of insurgent winners includes Kennedy, Nixon 1968, Carter 1976, and Bush 2000. JFK was sort of a once in a lifetime charismatic candidate, Nixon benefited from infighting by the opposing party, Carter benefited from Nixon’s scandals, and Bush benefited from Clinton’s scandals (although he may have been helped more by Gore’s distancing himself from a still-popular incumbent). In a past world, Harris may have benefited from Trump’s convictions and other scandals, but political scandals no longer have the same effect in the modern day. Running a strong insurgent campaign may have not been possible without more domestic policy issues that could galvanize the Democratic base and persuade undecided voters.