Post Election Campaign Narrative

About New Hampshire

New Hampshire, known as the Live Free or Die state, was one of the original 13 colonies of the United States. It has a small international border with Canada in the north and borders Massachusetts, Maine, Vermont, and the Atlantic Ocean on other sides. New Hampshire is known for its nature, including the highest peak in the northeast and more than 1300 lakes and 40 rivers, and many threatened animal species. Other large geographic features include the White Mountains in the north and Lake Winnipesaukee, the state’s largest lake. According to the New Hampshire Department of Business and Economic Affairs, its main industries are advanced manufacturing, technology, life sciences, and tourism and hospitality. By real value added to the GDP of the state in 2023, the largest industry was finance, insurance, real estate, rental, and leasing, followed by professional and business services.

The state has a population of 1,377,529 as of the 2020 Census. New Hampshire has two members of the US Senate and two members of the US House. The 2023 American Community Survey estimated that 40.7% of residents have a bachelor’s degree or higher and the employment rate is 64.7% (higher than US average by 4 points). Compared the US median age measured by this survey of 39.2, New Hampshire is older on average, with a median age of 43.4. The three largest age groups in the state are 50 to 54, 55 to 59 and 60 to 64. Given how we have seen in past weeks that this age group’s general lean is slightly towards Trump, it would be interesting to look at the demographic voting data if it is available. New Hampshire’s older population (65 years or older) is also larger than the US average. More demographics will be explored as we break down the results by county and compare them to predictions below.

New Hampshire as a Swing State

In the 2024 presidential election, New Hampshire was not considered one of the seven toss up states in the few months leading up to the election, instead considered to be leaning to the democrats. Since 2000, New Hampshire has voted Democratic 85.7% and Republican 14.3% of the time. The New Hampshire winners since 2000 are below:

• 2000: George W Bush
• 2004: John Kerry
• 2008: Barack Obama
• 2012: Barack Obama
• 2016: Hillary Clinton
• 2020: Joe Biden
• 2024: Kamala Harris

The state last voted for a Republican in 2000, and since has voted for the Democrat in each election, being incorrect in 2004, 2016, and 2024 in consistency with the overall winner of the electoral college.

In 2020, local news described the state as a battleground state that could help determine the outcome of the election. Reporting by Slate in 2020 found that while 2016 was the second smallest margin of all states, New Hampshire may be moving away from the swing state moniker it has had since the 1990s. Given it has gone blue every election since 2000, this may be the case, but these results have all been by incredibly slim margins, meaning it does not seem New Hampshire is quite set on being blue for the foreseeable future. This can also be seen by down ballot results and campaign involvement in the states. New Hampshire remains an up for grabs, tight margin state.

The 2024 New Hampshire Ballot

In 2024, the November 5 ballot included votes for the president, US House, governor, state executives, State Senate, State House, special state legislative candidates, and ballot measures.

The US House races both went to the Democrats, with Chris Pappas getting 54% of the vote in district 1 and Maggie Goodlander getting 52.9% of the vote in district 2 according to the Associated Press.

States races for the New Hampshire House of Representatives included elections for all 400 seats. The Republicans had the majority both before and after the election, and within state offices, the new State House, State Senate, and governor are now all Republican. There are 23 other states that also had this so called Republican trifecta. Before the election in the State House, Politico had reported that it was likely to flip to Democratic control, a prediction that did not pan out.

The election for the governor of New Hampshire was also quite close according to Ballotpedia. The Republican Kelly Ayote won against the Democrat Joyce Craig, and a third candidate for the Libertarian Party also ran. The primaries for this role took place on September 10, making the campaign between these two candidates head to head very short in what seems to be a theme of this election. Expert predictions did not pick much of a side in how the election would go, with The Cook Political Report and Inside Elections calling it a Toss-Up and Sabato’s Crystal Ball saying it would be Lean Republican. Both candidates were not incumbents as they were running to replace Governor Chris Sununu. The governor’s race had nationwide attention, with nationwide governors association pouring in millions, 2 million dollars from Republicans and 9 million from Democrats, on the day of the primary.

The state went to Kamala Harris in the presidential result and democrats for the US House seats but Republicans in all three of the state wide office results, leading to an interesting puzzle to determine whether ticket splitting or some other campaign effect is the cause of this. Additionally, it signals that New Hampshire is not clearly in the pocket of either party going forward and will continue to be a hub of spending and campaigning in future elections.

Presidential Election Predictions in New Hampshire

As New Hampshire was not one of the seven swing states I identified at the beginning of the semester based on expert predictions by Sabato’s Crystal Ball and The Cook Political Report, I did not calculate a personal prediction with either of my final models. To then assess how close predictions were to the actual results discussed in the first paragraph of this blog post, I will use the predictions from expert sources for comparison. These predictions and their methodology, as much as it is available, are below.

Starting with Real Clear Politics, which takes an unweighted average of polling results and as seen in the post election reflection was closer to the actual national result than FiveThirtyEight’s weighted approach, they predicted that New Hampshire would go to Harris 50.0% to Trump’s 45.8%, which a Harris+4.2 margin prediction.

For FiveThirtyEight, their final prediction as of November 3 was Harris 50.3% to Trump’s 45.5%, a 4.9 percentage point margin for the Democrats. This is a larger margin and higher Harris vote share prediction than given by Real Clear Politics’ final prediction. FiveThirtyEight began providing the Harris-Trump split for New Hampshire on August 21, and Harris never dipped below 50% at any time in the weighted average of polls based on their methodology and other factors as has been discussed extensively in past blog posts. Harris remained around the 50% number throughout the fall, which Trump’s prediction increased by nearly 3 percentage points to election day.

Sabato’s Crystal Ball final predictions did not provide popular vote breakdowns for presidential outcomes, but did rate New Hampshire as likely democrat They did mention that the gubernatorial race, saying it learned Republican because of the support from Governor Sununu for the Republican candidate.

The Cook Political Report also did not provide popular vote predictions but also put New Hampshire and its four electoral votes in the likely Democrat rating level.

Inside Elections rated New Hampshire Lean Democratic, which is the second category of moving in the democratic direction after Tilt Democratic. They also do not provide a vote share prediction.

Analysis of Expert Prediction Closeness

Based on these expert predictions, all of those who provided qualitative ratings were correct in the direction of their prediction. The two with popular vote percentage predictions were RCP and 538. RCP had Harris 50.0% to Trump’s 45.8%, a Harris+4.2 margin. 538 had Harris 50.3% to Trump’s 45.5%, a Harris+4.9 margin. In order to compare these to the actual result– Harris 50.9% to Trump’s 48.1%, a Harris+1.8 margin– I will look at comparing the margins and the popular vote outcome. As I did with my own predictions, I will look at a variety of measures of accuracy including the MSE, RMSE, MAE, and bias.

I will calculate all four of these measures and display them in the table below. The mean squared error is found by calculating the average of the squares of the errors. Because there are only two predictions to input, it will just be the average of those two’s squared errors. This value gives a sense of how large the errors tend to be, but the interpretation is abstract because it’s squared. Therefore, I look at the root mean squared error, or RMSE. Mean absolute error (MAE) measures the average of the absolute differences between predicted and actual vote share values. Like RMSE, MAE reflects the average prediction error, but unlike MSE and RMSE, it does not give more weight to larger errors. I also calculate each of these for the Democratic and Republican vote share because we are not using two-party vote share numbers. The margin is also used.

The MSE on the Democratic vote share value of 0.585 means that, on average, the squared error between the two-party democratic popular vote share the experts predicted and the actual outcome in vote share is about 0.585 percentage points squared. The predictions from experts’ error in predicting the Democratic vote share is relatively low, indicating good performance, but because of the difficulty of interpreting this I look at the RMSE. The low RMSE value of 0.765 aligns with the low MSE, confirming that NH Democratic vote predictions were relatively accurate. This can be interpreted as, on average, the predictions for the Democratic two-party vote share are off by about 0.765 percentage points. RMSE of 0.765 percentage points is a small error in the context where vote shares typically range from 40% to 60%. The MAE for democratic vote share is 0.75, meaning, on average, predictions for Democratic vote share are off by 0.75 percentage points. The bias helps us determine the direction of this MAE. With a bias of -0.75, the predictions seem to systematically underestimate Democratic vote share by 0.75 points. This is not necessarily consistent with the general trend across the country from the last election as was seen in the last blog post. This could be due to the low number of predictions being accounted. Overall, expert predictors did relatively well in predicting the Democratic vote share in new Hampshire, though they slightly underestimated it.

The results for the Republican vote share provide context for this negative bias which points to the underestimation of democrats as it shows that republicans were underestimated even more than the democrats in New Hampshire, making this more consistent with national trends of republican gains across the board. Looking at each of the evaluation metrics, the error is much higher for Republican predictions, suggesting the experts less accurately estimate Republican vote shares. This is consistent with the narrative of republicans under performing in polls and expert predictions. The MSE of the republican vote share is 6.025 squared percentage points, which is difficult to interpret, so using RMSE we see that the predictions are off by 2.455 percentage points on average. The MAE says that Republican vote share deviate by 2.45 points, much larger than the Democratic MAE, suggesting room for improvement in predicting Republicans accurately. The bias is the negative of this, meaning the experts underestimates Republican vote share by a margin of 2.45 points. The fact that this is much larger than the margin for Democratic vote share indicates a systematic issue with predicting Republican vote values for these two expert predictions, at least in New Hampshire, but this is a pattern that has been recognized across the board in the last three elections.

Looking at the margins may provide a way to reconcile this under-prediction of both Democrats and Republicans in models of popular vote in New Hampshire. It seems the models left much more room for third parties than actually occurred though, leaving margins also not very accurate. As seen in the table above, margin predictions suffer due to compounded errors from both party predictions. With an RMSE of 2.772, the margin prediction on average was off by 2.772 percentage points, enough to mispredict the election result significantly especially in such a close state. The MAE and bias are the same, indicating an overall overestimation of Democratic advantages in the margin. The positive bias in margin indicates the experts tended to overestimate the margin in favor of the Democrats, consistent with national trends.

Review of the 2024 Presidential Campaign in NH

As a part of my weekly predictions, I have also been following the news coming out of New Hampshire about the presidential and other campaigns. As a result, in the below section I will summarize how the campaigns did in the state over time and what the reaction was. In this analysis, I will consider how people were reacting to the quality of each candidate, the volume of campaign activity in the state at each time, what each campaign chose to talk about in the state, and how the news interpreted each action.

My following of these campaigns begins on September 17, just one week after the primaries for the New Hampshire governor had occurred. On this day, the top stories included one on comments made by the New Hampshire Libertarian Party about Harris. The party was running a gubernatorial candidate in NH, but the presidential Libertarian candidate said he thought the statement was abhorrent. It describes people who would attempt to assassinate VP Harris as an American hero, posted weeks after the first assassination attempt of Trump and the same day as the second at a Florida golf course. Instead of backing down, the Libertarian Party in New Hampshire attacked their own presidential candidate as an “infiltrating leftist snake” and used slurs in their response. A story this same day by the New Hampshire Bulletin indicated that the two major candidates, Trump and Harris, and their campaigns were not spending much of their time discussing the assassination attempt. Instead, both campaigns were hosting events in the state, with the Harris campaign focusing on her vision in contrast with Trump’s Project 2025.

Three days later, on September 19, stories focused on the new issues being focused on by voters and the campaigns in the state. The Portland, ME based television station WGME talked to New Hampshire voters who focused their concerns on the economy and immigration. These are the two areas where in the post election coverage, Trump’s rhetoric appeared to have been better. On the economy, many voters talked about inflation and how expensive food was. And despite the distance from the Southern Border, NH voters were also concerned about immigration. However, a UNH survey found 43 percent of residents believe illegal immigration is a very to somewhat serious concern for New Hampshire, a number that is much lower than the 83 percent nationally.

Additionally, around this time the Republicans launched a paid media campaign with ads in New Hampshire, Arizona, and Wisconsin against local Democrats who they called the “Left’s Most Wanted.” This was admist the release of a UNH poll that had Harris up 11 points in the state, which was much larger than the margin in the end. New laws in New Hampshire this week were proposed that would require people voting for the first time to bring documentation of citizenship. This would not effect the current election but was already challenged in court after its signing due to its potential to confuse and dissuade people from voting. Common news stories also revolved around the governor’s race, as Democrats indicated they were optimistic in potentially flipping the state house to a Democratic governor.

For the rest of September, a few stories stuck out on voting access, top issues brought forward by the candidates, and the idea of ticket splitting. First, on the way that voting would be carried out in the election, stories in local newspapers began to focus on how voters can and cannot vote in the state. New Hampshire does not offer early in-person voting but does offer absentee ballots under certain specific reasons. Residents also have to register between 6-13 days before election day across the state.

Stories about ticket splitting also emerged. In New Hampshire, the Harris campaign told News 9 it saw an increase Republicans registering to volunteer with their statewide coordinated campaign by 47%. This includes joining the door knocking efforts across the state, one of the most effective ground campaign methods. The New Hampshire ground game has always played a significant part in the election. Polls from UNH are also described to have found that more 2020 Democratic voters were willing to support the Republican governor candidate versus 2020 Republican voters going the other way, pointing for a potential to split the results by party between the governor and president elected in the state. This is ultimately what happened. CNN also reported that the former state GOP chairman noted that if Harris is only able to win by a few points, the Republicans could still hold the governorship due to ticket splitting, which ultimately occurred.

The messages throughout September also played a large role in news coverage. The role of labor unions as reported by the New Hampshire Bulletin was that they were working to engage their voters in the process and encourage them to vote. Trump also introduced new rhetoric around this time as reported by New Hampshire’s NPR branch. He focused on being tough on crime and harsh attacks on democrats, including Harris and Biden.

The news continued to pile up through October as the campaigns remained highly involved in the New Hampshire election. It appears both candidates themselves didn’t return to the state after September, but their campaigns and ad spending remained high. For Harris, the coordinated campaign of events and on the ground volunteers was emphasized as they focused on ground game.

In local elections, candidate forums were planned across the state, also likely to allow voters to discuss their preferences and bring up the presidential election in the discussions. Many Massachusetts Democrats also chose to get involved in the New Hampshire elections, crossing the border to get involved in the ground campaign.

A key concern in NH included housing and how expensive and inaccessible it was, but the New Hampshire Association of Realtors announced they would not endorse any candidates in October, instead continuing to push for their issues and that they were happy they were being discussed.

In order to reach more voters, the New Hampshire Bulletin noted that the candidates were turning to TV, podcasts, and radio. This focus on the air war with the top of the ticket coincided with on the ground campaign work with guest visitors and volunteers. On October 8th, Senator Cory Booker campaigned for Harris in the state. Gwen Walz was the next to visit in the second to last week of October, followed by an announcement that President Biden and Bernie Sanders were to visit a community college in Concord for the campaign. A visit from Bill Clinton was also planned for early November.

In October, discussions of spending were also common, with one report noting that nearly 15 million dollars was spent on the governor’s race advertising alone (10 million by Republicans, 4.5 for Democrats). In addition to spending, state leaders were also working on endorsements, and two NH Republican Party leaders endorsed Harris in late October.

Polls in late October were also incredibly close, with some indicating Trump could win and others maintaining a small margin for Harris as had been the case for the past months. On average, Harris was still favored.

In the final days before the campaign, JD Vance traveled to New Hampshire for one final rally there, closing out the New Hampshire campaigns for the parties on November 3rd. Ads and ground game would continue through election day, eventually leading to a Harris victory.

Campaign and Candidate Effects As Deviations from Predictions

The deviations from the forecasted outcome showed that Republicans were systematically under-predicted, as was true across the country. As seen in Gelman and King (1993), the polls should become more accurate towards the time of the election becuase people start to pay attention and learn about the candidates, but this still did not fix the final predictions being off.

In Lynn Vavreck’s 2009 book “The Message Matters”, she defines the idea of how candidates running the right type of campaign can outperform fundamentals, splitting candidates into clarifying and insurgent campaigns based on who the economy helps. In class, we defined Harris as the insurgent campaign in this cycle because of her being a representative of the incumbent party in a “bad economy.” Her goal would then be to talk about beliefs outside of the economy which benefit them in terms of public opinion, while the clarifying candidate like Trymp would focus on the economy over anything else. Based on the news coverage discussed above, Harris and her surrogates spent a lot of time in New Hampshire talking about abortion, her vision for America, and other domestic policy issues while Trump and team focused on the economy, which would confirm the Vavreck theory occurring. This would allow Trump to overperform the predictions of his vote share as we saw he did, but the same for Harris, though Trump did so more. This could be one example for why both overperformed predictions but Trump more.

The news articles also highlighted the importance of the ground game to the two campaigns. Kalla and Broockman’s 2018 paper “The minimal persuasive effects of campaign contact in general elections: Evidence from 49 field experiments” found that across many field experiments, there was no real impacts on persuasion of voters as a result of ground campaign efforts. This meta analysis of other studies looked at the general election outcomes and saw overall no significant differences from 0. There are some other impacts of the ground game, though, as shown by Enos and Fowler in 2016. This does not provide an exact reason for why more voters were persuaded to Trump proportionally to Harris (in addition to general polling unable to capture Trump support), but an uneven increase in turnout for Republicans could explain the deviation. The Republican campaign ground game efforts could have led to an increased turnout than expected, leading to that increase.

Many articles also focused on ad spending and the Air War. In Huber and Arceneaux’s “Identifying the Persuasive Effects of Presidential Advertising,” the study finds that presidential campaign advertisements do not significantly mobilize or inform voters but are highly effective at persuading them to support a specific candidate. They indicate that the persuasive effects of presidential advertisements can be significant but are generally short-lived. The impact is most pronounced right after exposure and diminishes over time unless reinforced by more advertising. A similar finding is revealed from Geber et al’s “How Large and Long-lasting Are the Persuasive Effects of Televised Campaign Ads? Results from a Randomized Field Experiment,” which says that TV ads have strong immediate effects on voter preferences, with a maximum boost occurring during the week of exposure, but effects decay rapidly, becoming negligible within one to two weeks.

All in all, the deviations between predictions and the actual results in New Hampshire can be partially explained by campaign effects, but likely have inherent biases as well.

Thank you for following along these past few weeks to my forecasts and post election analyses. JS.