
For reference, all election ratings used in the text and maps of the article employ this color-coded system:
Safe Democratic: Dark Blue
Likely Democratic: Light Blue
Lean Democratic: Pale Blue
Tossup: Beige
Lean Republican: Pale Red
Likely Republican: Light Red
Safe Republican: Dark Red
All polling averages are from FiveThirtyEight.
Ratings Changes
Wisconsin (President): Lean Democratic -> Tossup
Michigan (President): Lean Democratic -> Tossup
Montana (Senate): Lean Republican -> Likely Republican
The Presidential Election
NATIONAL: Harris +2.4 (+0.3)
Swing States (all Tossups)
Wisconsin: Harris + 0.6 (+1.1)
Michigan: Harris + 0.6 (+1.0)
Pennsylvania: Harris + 0.6 (+0.2)
North Carolina: Trump +0.9 (+0.4)
Georgia: Trump +0.9 (+0.3)
Arizona: Trump +1.4 (+0.2)
Nevada: Harris +0.7 (+0.4)
Peripheral States
Florida (Likely R): Trump +4.9 (+1.1)
Texas (Likely R): Trump +6.7 (+0.5)
New Hampshire (Likely D): Harris +6.6 (+0.5)
Virginia (Likely D): Harris +7.6 (+0.7)
The State of the Senate
Key Seats
Michigan (Lean D): Slotkin +4.0 (+0.7)
Ohio (Tossup): Brown +2.4 (+0.3)
Texas (Lean R): Cruz +3.7 (+0.3)
Montana (Likely R): Sheehy +5.4 (+1.1)
Florida (Likely R): Scott +4.5 (+0.1)
Nebraska (Likely R): Fischer +2.0 (+0.6)
Gubernatorial Elections
Weekly Recap: Natural Disaster Politics and the “Herding” Phenomenon
The past few weeks of the 2024 Presidential campaign have proved to be a bit of a slog. Kamala Harris’s initial surge post-Biden-dropout appears to have come to an end, and the race has settled into a coin-toss. Neither candidate seems to possess lasting strength in any of the seven major swing states, and Harris’s consistent but relatively small lead in the popular vote has not changed much. The picture in the Senate is equally murky; while Republicans seem to slowly be eating away at Democrats’ already-low chances of holding onto the Senate, Republicans will probably not flip enough Senate seats to secure any lasting dominance in the chamber. Democrats, meanwhile, are stuck in a sort of limbo; there are a few Republican Senate seats they could target, but no particular one that stands out as most likely to flip. In the House of Representatives, there is no clear favorite—though, unless the election is very close, whoever wins the Presidency is more likely than not to win the House.
In this environment of close poll after close poll and no clear favorite in sight, it is unsurprising that pundits have been grasping for an “October Surprise” that will completely upend the race, as discussed in the previous article. Hurricane Milton, which recently swept over Florida and killed more than a dozen people, was discussed as having the potential to change the race—and it is entirely possible that it could, though possibly not in the way one would expect. Meanwhile, a phenomenon in polling known as “herding” could explain why there seems to be an endless succession of razor-thin polling margins in this race, and why the polls may not be as trustworthy as they seem. This week’s recap will discuss both of these possibilities, and how history and statistical analysis can help us understand a highly uncertain election.
Hurricane Katrina and the Political Effect of Natural Disasters
Conventional wisdom holds that unless the government response is universally acclaimed, natural disasters—beyond the immense suffering they cause to the areas that experience them—tend to be bad news for people in power. For most Americans, Hurricane Katrina serves as a textbook example: New Orleans and large swathes of Louisiana and Mississippi were devastated by the hurricane, over one million people were displaced, and both local governments in Louisiana and then-President George W. Bush faced intense criticism for their handling of the crisis. One would expect, then, that the Republican Party would have suffered in the region when the next Presidential election took place in 2008, as angry residents took the chance to use the power of the ballot against a governing party that had failed them.
On the national level, Republicans did suffer: Barack Obama won handily against Republican candidate John McCain as Democrats swept the House and Senate, in large part because of outgoing incumbent George Bush’s unpopularity. In Louisiana, however, McCain defied the trend: despite expectations that he might, as a Republican, suffer from ties to Bush’s response to Katrina, he actually did better than Bush’s 2004 result in the state. Louisiana was not the only state in which McCain outperformed Bush—he also did better in Oklahoma, Arkansas, Tennessee, and West Virginia. Analysts have explained the shift in the latter four states as a result of disenchantment with the Democratic Party among their deeply conservative and religious inhabitants—and because residents who voted for Hillary Clinton over Barack Obama in that year’s Democratic primary election switched to the Republican Party for the general election.
In Louisiana, this was not the case: while religious conservatism is also predominant in Louisiana, it was also one of Barack Obama’s strongest states in the Democratic primaries. Despite this, many parishes that Obama won handily in the primary still swung right—parishes that also often happened to fall within the areas of Louisiana most impacted by Hurricane Katrina. Plaquemines Parish in the coastal Southeast, for instance, voted for Obama in the primary but swung right in the general election. St. Bernard Parish, which lost over half of its population after the hurricane, swung over 10 points to the right. The pattern did not just extend to Louisiana. In Mississippi, two of the few counties that swung right in the state were located just across the Louisiana border, in the area of the state most impacted by Katrina.
Thankfully, the devastation Hurricane Milton wreaked on Florida does not seem to have reached the scale of Hurricane Katrina. What remains to be seen, however, is how effectively the government responds to the disaster. A poor response could lead to weeks of negative press coverage for Democrats. Regardless of how the response is received, however, observers should keep an eye out for what the results of the election look like in the areas of Florida most impacted by the storm—it may not be enough to impact the overall result, but it will not be surprising if Harris sees a slight boost in votes in those areas (or at least will lose fewer votes than in the rest of the state, given Florida’s recent rightward turn).
The “Herding” Phenomenon
If one were to look at recent polling in swing states, it is abundantly clear that results of most polls show a close race. This is to be expected in swing states, but the more one looks, the more it seems that the polling results are almost too close. Looking at FiveThirtyEight’s polling averages, there is only one swing state out of seven where even a single recent poll shows either Trump or Harris with a lead above five points: Georgia. Even within Georgia, there are only two such polls, and one of them is from a pollster funded by the Trump campaign.
To understand why this is strange, it is necessary to understand some principles of statistics. In a normal statistical environment, there are values that fall close to a mean, and there are outliers. To put this into the context of swing state polling, this means that in any given swing state, one would expect to see many polls that are close to an overall average, and a few scattered outlier polls (if you lined up all of the polls by their margin, it would look something like a bell curve). The polls close to an average are certainly present, but what is odd is that there are very few—indeed, almost no—outliers.
This could be explained by a phenomenon known as “herding”. In essence, herding is a practice in which pollsters, worried about the accuracy of their polls, tweak the results or shift their methodology to ensure the final numbers are relatively close to what other pollsters are showing to avoid being singled out for blame if polling is inaccurate. When herding occurs, there are very few outliers and most polls fall close to the mean: exactly what we are seeing in swing states now.
This is where things get hazy, since pollsters are rarely this open about their methodology—especially when that involves disclosing that they may be “fudging the numbers.” But it might make sense: many pollsters, having suffered misses in both the 2016 and 2020 elections, are absolutely terrified that another miss in 2024 could destroy their credibility for good. For these pollsters, the safest bet is to simply release polls that indicate vaguely that “the race is close”—after all, one candidate has to win, and it’s always better to simply say the election is a tossup than to face the humiliation of saying a candidate is favored when they end up losing. From a pragmatic standpoint, this is understandable: pollsters are run by people with jobs, and it is better for their livelihood to not stand out from the crowd. From an electoral standpoint, however, it can be misleading: campaigns use polls to determine funding, and voters use polls to understand which way an election is going. If the election turns out to not be close—regardless of who wins—herding might offer us an explanation for why polling didn’t reflect it.
The opinions expressed within this piece represent the views of the author alone and do not necessarily reflect the views of The Jefferson Independent.
Leave a Reply