November 6th, 2024
Wow, I think it is safe to say almost all the polls were wrong and biased toward error for Harris!
I'll take a little break and let the dust settle, but a quick look at yesterday's poll data shows Trump performing better in almost all polls, national or battleground state. Only one polling firm was within a percentage point of all national and state results, the Brazilian firm, Atlas Intel.
I will conduct a deeper analysis of what this means for the other polling firms. But a national news organization, CNN, was off on its Wisconsin poll outside of its huge +/-4.8 margin of error. And this poll probably contributed to the RCP average for Wisconsin flipping for Harris for a state that Trump would eventually win.
November 5th, 2024
Hello everyone. And Happy Voting Day!
The polls will officially fade in importance after today. So go out and be a likely voter!
For your viewing pleasure, I've captured many of the relevant polling data from realclearpolitics.com. When the results are in you can see for yourself just how accurate or inaccurate the polls were.
Enjoy!
https://www.realclearpolling.com/maps/president/2024/no-toss-up/electoral-college
https://www.realclearpolling.com/polls/president/general/2024/trump-vs-harris
https://www.realclearpolling.com/polls/president/general/2024/arizona/trump-vs-harris
https://www.realclearpolling.com/polls/president/general/2024/georgia/trump-vs-harris
https://www.realclearpolling.com/polls/president/general/2024/michigan/trump-vs-harris
https://www.realclearpolling.com/polls/president/general/2024/nevada/trump-vs-harris
https://www.realclearpolling.com/polls/president/general/2024/north-carolina/trump-vs-harris
https://www.realclearpolling.com/polls/president/general/2024/pennsylvania/trump-vs-harris
https://www.realclearpolling.com/polls/president/general/2024/wisconsin/trump-vs-harris
https://www.realclearpolling.com/betting-odds/2024/president
November 2nd, 2024
Here we are the weekend before election day. Many pollsters have released their last polls. And the RCP Averages have Harris regaining the advantage in two swing states, Wisconsin and Michigan. However the RCP betting average still has Trump favored 58.2-40.7.
I have already covered how inaccurate polls are because they have built in large margins of error. And margins of error only relate to random sampling error. Polls can still be subject to all sorts of errors, including intentionally biased questions. All of this inaccuracy and bias allows pollsters and news organizations to use polling as false cover for their own favored politcal narratives.
So what should we do now that the poll differences between candidates are so buried in their statistical insignificance? And so many pollsters and news organizations are calling the race a coin flip?
The truth is, thanks to COVID measures still practiced in most states, many people have already voted. So we actually don't need polls anymore to gain some insight in to how the election is shaping up because we have a steady stream of voters both mailing in their ballots and voting in person before election day. NBC has early voting results for 2024 so far and we can compare the results by party registration to the total early voting results in 2020.
The first six bars show the progression support for each party from 2020-2024. Blue - Democrats, Red - Republicans, and Grey/Black - Independents. The last two bars show the margin of support for Trump, first for the actual vote in 2020, followed by the RCP State Average. Below 0% means Trump lost or is losing the state. And above 0% means he won or is winning the state. These numbers are small because in both 2020 and in 2024 these states are showing very small margins of difference between the two candidates. Trump lost all these states except for North Carolina in 2020. And he is ahead in all these states except for Michigan and Wisconsin in the RCP averages.
There are some important things to note before I go over the revealed patterns.
1. The early voter data is for party registration only. It does not indicate who the voter actually voted for.
2. There is a noted shift for Trump in Black and Latino communities. Those two groups who previously favored Democrats have diminished or flipped in their support for Harris. It is unclear if previously registered Democrat Trump supporters have changed their registration status before they voted or if they are still voting as Democrats. Arab American registered Democrats who will either support Trump or vote Third Party are also not included as Republicans or Independents. This is especially important in Michigan with its large Arab American population which is actively protesting Harris' military support of Israel.
3. Pennsylvania in-person mail in voting was recently extended because of an appearant heavy turnout in Republican areas. The heavy turnout resulted in long lines and reports of voters being turned away. So the deadline was extended.
4. The only polls not showing Republican gains and Democrat declines (Georgia, Michigan, Wisconsin) do not report early vote by party registration. The data for these three states are statistically modeled. And this modeling is subject to the same bias and sampling errors as already discussed for polls.
The first three states, Nevada, Arizona, and North Carolina show very straightforward early voting patterns. There is a decrease in Democrats and an increase in Republicans. And the RCP Poll Averages show a shift toward Trump from 2020. This is consistent with a fundamental shift in both turnout and actual voting preferences for Trump. These three states would appear to be very unlikely to shift toward Harris on election day. Especially since Democrats have historically been better at turning out the early vote with Republicans coming out on election day.
Georgia shows an increase in Democrats and a decrease in Republicans. However, the RCP Average is the largest for Trump of all seven states. So this still points to a fundamental shift in voting patterns towards Trump. Perhaps more disaffected Republicans are coming home to vote for Trump. The Republicans still have their turnout edge. So if the Republicans really are coming home, Trump should win with a strong showing in Georgia. This means if Trump can take just one of the remaining "blue wall" states he wins the presidency.
Pennsylvania shows an even larger drop in Democrats and a rise in Republicans. If the Republicans come out strong on election day (same as their long lines for Republican early vote centers) then Trump will win Pennsylvania and the presidency. The Trump campaign has to feel good about their chances.
The other two states' early voting performance is less straightforward. Both parties increased their participation from 2020 in Michigan. However the Democrats' increase is greater. Is this because of the turnout of Arab Americans and other groups still registered as Democrats but favoring Trump? The answer to this question could very well determine if Trump wins Michigan. Wisconsin has a severe decline in Republicans. But there is a dramatic increase in Independents. Are the Independents breaking for Trump in an anti-incumbent year? This is unknown. But if Trump wins the other five states, these last two would only be important for increasing Trump's electoral college victory.
The early voting registration numbers look very good for Trump. Based on the data presented here, Harris' only chance for victory appears to be sweeping the three "blue wall" states.
October 28th, 2024
With just a week to go until election day, Trump has taken a slight lead in the RCP National Average. Trump now leads Harris 48.5%-48.4%. Trump's steady climb and Harris' steady decline now has Trump leading in the national average and each of seven key battleground state averages that Realclearpolitics tracks - Arizona, Nevada, Wisconsin, Michigan, Pennsylvania, North Carolina, and Georgia.
The steady movement in only one direction - positive for Trump - can't be explained by random movement within a margin of error confidence interval. Potential voters really appear to be breaking for Trump in increasing numbers. Analyzing individual polls at this point is less useful than looking at this consistent improvement each day as new polls are released and averaged in. The margins of error of each individual poll are just too large to say anything with statistical significance from each poll. However, the lack of random fluctuation within the confidence intervals from all the polls is important as it shows a clear trend as the horserace shifts towards Trump.
The RCP betting average continues to reach new highs for Trump. It currently favors Trump 61.1%- 37.5%. And anecdotal headline stories served on Realclearpolitics.com show an increase in Democrat panic stories.
October 25th, 2024
It appears that Harris has now lost her lead in the national polls. The RCP National Average now has Trump and Harris tied at 48.5%. And new polls by both New York Times/Siena (48%) and CNN (47%) have Trump and Harris tied nationally among Likely Voters.
This is bad news for Harris as this continues her steady erosion of support in polling over the past few weeks. And with polls biased in favor of democrats in the previous two presidential election cycles, it is likely that Trump is actually ahead in the national vote as well as the battleground states.
The Harris camp appears to be taking this drop in support seriously and is looking increasingly desperate as they try and retool their campaign in the final days. The joy is gone as Harris and her surrogates are openly calling Trump a fascist. This is despite the elevated threat to Trump's life such language causes even after two recent assassination attempts.
October 23rd, 2024
Another poll metric moving in Trump's direction. Gallup released a poll on candidate favorability. In a poll sampling national adults, Trump now is +2 in favorability 50-48%. Harris is now -2 in favorability 48-50%.
This is just another of the continuous march of new poll numbers which have Trump increasing support and Harris decreasing support among the American people.
This poll is particularly interesting because it polls all adults, not registered or likely voters. So an obvious question to ask is if Trump is such an existential threat to democracy, why would a majority of American adults view him favorably? Democrat messaging would appear to have a credibility problem. Poll after poll is showing a shift towards Trump with only two weeks to go until election day. And millions of votes are currently being cast every day in early voting.
October 22nd, 2024
Buried in the new USA Today/Suffolk poll, in it's report titled, "Exclusive poll: Harris and Trump tied amid battle for Latino and Black voters", is the fact that the poll has Lantino voters favoring Trump 49%-38%. The report dismisses this huge margin for Trump as a tie by saying, "The margins of error for the small subsamples of Latinos and Black voters are plus or [sic] 9 points − a potential shift of up to 18 points one way or the other − and other recent polls show Harris in a stronger position, including a lead among Hispanics." This is a rather odd criticism of their own poll numbers, when as mentioned consistently in this blog, almost all of the margins of error in election polls are huge for the small differences they want to measure between candidates. This result is no different from other results like who is ahead in the popular vote or in battleground states. So why not just say most polls are too inaccurate by design to say much of anything with statistical significance. We're now going to pontificate on whatever we want and still use the poll as evidence.
However, this is just another notable shift towards Trump among all the other shifts in the last few weeks that can't be due to random sampling error if they only move one way. + or - has to mean movement in both directions randomly, not just ++ for Trump. A trend is definitely emerging of increasing support for Trump across different voting groups.
Harris' advantage in the latest national RCP average is just +.8. When Biden at this point in time had a +7.9 advantage and barely won. And Clinton had a +6.1 advantage and barely lost.
The betting markets are looking at how far Harris has fallen and are betting more against her every day. The latest RCP Betting Average has Trump at 60.1 and Harris at 38.4. A long and steady climb for Trump that has him almost at the 66 or so percent chance of winning that he enjoyed at the height of Biden being forced out of the race by the Democrat party.
October 21th, 2024
Harris continues to slide in the RCP National Average with Harris up just 1.0 points. It briefly dipped below 1 before a +1 poll for Harris from USA Today/Suffolk was averaged in.
This slim 1.0 point advantage for Harris exists only because polls from early October are still being averaged in. The majority of polls released over the last week or so (when early voting is actually occuring) show Trump leading in the national samples. As mentioned before, if this movement was due to random sampling error related to the large margin of error in each poll, the polls should be moving randomly in each candidate's direction within the confidence interval. That polls almost exclusively show a movement towards Trump is important.
The betting averages continue to improve for Trump. The RCP Betting Average now has Trump up 59.0 to 39.8. Three betting firms have Trump at 60-61 and three have Harris in the high 30s. The betting firms are clearly showing that there is little confidence in the accuracy of individual polls and poll averages. They seem to support that the polls are heavily biased for Harris and the actual numbers favor Trump. Since Trump is only up 1.0 points in the RCP Battleground Average as well, the nearly 20 point spread in Trump's favor from the betting firms is in contradiction to the slim margins in the actual polls.
October 17th, 2024
More movement away from Harris. the RCP Battleground Averages have Trump up 312-226 in the No Toss Up map of electoral votes. Trump now has a lead in all 7 battleground states in the RCP Averages.
This continued movement of the polls in just one direction, for Trump, appears to be a clear pattern and is not expected of large margins of error causing large sampling error jumps both for Trump AND Harris at random.
The betting markets have noticed. The RCP Betting Average now has Trump up 58.1 to 40.9. This is almost a 20 point spread. And it is larger than the margin Trump had just before Biden dropped out of the race. This probably reflects the short time for anything significant to happen before election day to change Trump's new momentum as the early vote takes place locking in the new Trump advantage.
October 16th, 2024
A Harvard Harris Poll has some interesting data on early voters. In the 2020 election, Election day voters favored Trump over Biden 65%-33%. Trump even won early in-person voting 52%-47%. But in the mail/absentee vote, Biden won by flipping the advantage to 65%-33% for Biden.
In the Harvard/Harris poll, among participants who already cast their vote in battleground states, Trump is up 48%-47%.
Since Democrats often take advantage of their large funding advantage by vote harvesting (where Democrat ground personnel go in to vote friendly areas and offer to hand deliver absentee ballots), this margin really needs to be closer to 60% Harris for her to stay in the race to combat Trump's in-person advantage. It will be interesting to see if this metric moves in Harris' direction closer to election day.
October 15th, 2024
The phrase is going around, "Margin of Error Election". While technically true, it's only a Margin of Error Election because pollsters design huge margins of error in their polls. As wide as the margins of error are, even a state like Texas has poll results that are within the margins of error. The latest Marist Poll with a margin of error of +/- 3.6 means that Harris could have support as high as 49.6% and Trump support could be as low as 49.4% in the margin of error confidence interval. Meaning the poll result of Trump +7 53%-46% is within the poll's margin of error and Texas is a Margin of Error Election. And no one really believes that Texas is a toss up state. The solution is simple, sample more people and decrease the confidence interval.
The latest Morning Consult National Poll has a margin of error of just +/- 1. Their result has Harris up +4 50%-46%. So the interval could have Harris as low as 49% and Trump as high as 47%. And this still gives the advantage to Harris. So it is outside the margin of error even with the smaller +4 to the larger Texas +7 difference between the candidates. Does that mean Harris is really ahead in the popular vote? No, it just means that with the Morning Consult results, we can stop using the misleading "Margin of Error Election" phrase in election discussions. Then we can focus on the real reason polls are often incorrect, intentional pollster bias.
The latest NBC National Poll has Harris and Trump at 48% each with a +/- 3.1 margin of error. This is a decline from the previous month's poll which had Harris up +5 49%-44%. In addition it shows Harris actually behind by a point when other candidates are included who could be on the ballot. And Harris dipped negative in popularity while Trump improved. But with such a large margin of error, this is another example of a "Margin of Error Election Poll". So with the headline that the election is deadlocked, poll watchers are distracted from the pollsters directly injecting their bias into the poll results through their faulty poll methodology.
The poll summary states, "While most public polls, including September's national NBC News poll, find the cost of living ranking as voters’ top concern, this survey asked a different question to get at voter intensity and motivation..." However, this poll shows that it does not prioritize voter intensity and motivation because the registered voter sample used in the poll includes a large number of unmotivated people who will never vote in this election cycle. If NBC pollsters were really interested in voter intensity and motivation, they would have used a likely voter sample which actually models voter intensity and motivation to show up on election day and vote. Registered voter samples often over sample groups that usually vote for the Democrat party. And it shows that NBC wants this bias in their poll (at the cost of not sampling the most motivated voters) when most other national pollsters switched to likely voter samples over a month ago.
So if voter intensity and motivation is not important to NBC, then why did they need to change the methodology of their poll and ask a different question? The question they asked is, "Is there one issue you feel so strongly about that you will vote for or against a candidate solely on that issue?" In short, they no longer wanted to poll voters who looked at a variety of issues and used their critical thinking skills to weigh candidate positions and vote based on many relevant voter issues like inflation, crime, illegal immigration, and war in Europe and the Middle East. NBC was now only interested in voters who were one issue ideologues who would ignore the totality of domestic and international challenges a President must solve and only vote based on a single issue. And it appears that the one issue that single issue ideologues vote for more than any other issue as far as registered voters go is abortion.
NBC News is not even pretending to be politically neutral anymore. They wanted to print the line, "Abortion is the top motivating issue - and the best issue for Harris" And they didn't care how much they had to twist their top issue question in to a pretzel to do it. Shame on NBC for such obvious bias in reporting. But in a nation where 60 minutes edits in multiple answers to the same Harris interview question and CBS and ABC only fact check the Republican debate candidates, I guess it isn't surprising that the major networks are just turning in to the propaganda arms of the Democrat party. As has already been stated in this blog, many pollsters and news organizations value political narrative over accuracy.
October 11th, 2024
Harris' poll numbers continue to slide with each new poll. Given that the polls still have huge margins of error, news outlets continue to call the race close and a toss up. But remember that margins of error mean that random sampling errror can give you different results within a confidence interval. So with each new poll released, you would expect some numbers to be larger than the last time the pollster sampled the same population and some numbers would be smaller just by random sampling variation. That is, if the real numbers haven't moved since the last time the sample was taken.
That's not what's happening,with each new poll, Harris drops a few percentage points and Trump gains a few. All the polls are moving in one direction, improving for Trump. And that can't be a result of random sampling error. That's a pattern of Harris sliding.
Depending on the news outlet, you'd never notice that Harris appears to be losing more and more battle ground states. The latest RCP average to flip for Trump is Nevada. Trump now leads 48.2-48. And he has a 302-236 advantage in the no toss up electoral count. The RCP poll averages have Trump winning 6 of 7 battlegrounds. Yes aggregate poll averages are poor measures. But as state after state flips, the polling is showing a clear pattern of momentum for Trump independent of poor statistical significance in individual polls.
The betting money favors Trump by an ever wider margin with the RCP Average at 53.9-44.7. Every odds maker favors Trump. With a high of 56 Trump to a low of 42 Harris. The best Harris can do is 47 to Trump's worst at 52. Smart money says this race is not too close to call or a toss up. Odds makers now think Trump is ahead. And each day Harris' slide gets worse. Trump has now surpassed Harris' peak in the RCP Betting Average. Remember how enthusiastic news organizations were when Harris was riding this peak just two months ago? You'd never know from most news outlets that now Trump has this same advantage just a few weeks out from election day.
October 9th, 2024
There is a new New York Times/Siena Poll. CBS News is making a big deal about it. They note that this is the first time that Harris has taken a lead in this national poll with Harris up +3. (Of course this is still within the margin of error and is statistically not a significant difference.) However this doesn't stop them from claiming that this means that voters are increasingly seeing Harris as the change candidate and the clear implication is that her momentum is building going in to the final few weeks of the election cycle.
However, internal polling done for both Democrats and Republicans shows Harris now losing in Wisconsin and also in trouble in Michigan. It is troubling that both the public poll and CBS narrative seem to be the exact opposite of the internal polling done by the political parties themselves. News organizations have a responsibility to the public of offering accurate information so that the American public can make good decisions based on good information. Are the public news organizations doing their jobs?
I'd say no, they might even be gaslighting the public at this point. As bad as the public polls are in accuracy with horrible margins of error, first Pennsylvania and now Michigan have flipped in the RCP averages. With Trump now leading Harris 296-242 in the electoral count. Wisconsin is only .5 points from flipping as well.
And when betting money is at stake, the RCP Average has Trump ahead 51.3-47.3. All three of those states are must win swing states for Harris. Yet there are still many stories in the news of her good national poll numbers and positive news cycle narratives.
October 7th, 2024
In the new Yahoo Poll.They state that Harris is leading Trump in the overall vote. But they have to qualify it, "Harris (48%) edges Trump (46%) by just two points. The new poll of 1,714 U.S. adults, which was conducted from Oct. 2 to 4, finds Harris with an even smaller advantage among registered voters when third-party candidates are included (46% to 45%). Among likely voters, the vice president’s lead vanishes entirely (47% to 47%)"
And if you've been paying attention to the contents of this blog, you should know by now that the only voters that count in predicting the election are likely voters. So the lead point of this poll is that the candidates are tied in the poll. The small two point lead among all adults may be the headline Yahoo wants, but the poll does not support it.
In addition, they say that the margin of error of the poll is 2.9%. But they don't break it out by subgroups for registered voters and likely voters so the margin of error should be higher for the subgroups. They eventually say because of this that the difference is too close to call.
But really, from the information in this blog you should now be able to say that the race is not too close to call, it's just that Yahoo's poll is too inaccurate to measure the difference between the candidates with statistical validity.
September 30th, 2024
First, a quick note on how to view margin of sampling errors. If you don't have time to read up on it in more detail, I'll give you a simple analogy. I've already said on this blog that the margins of errors for many election polls in the news are too wide to measure the small differences in a close election race with statistical significance.
Say you wanted to measure if a toothpick is 3 inches or 3 1/4 inches. But the only ruler you had was marked at zero and the one foot mark. That 12 inch interval between marks is just too wide to measure if a toothpick is 3 inches or 3 1/4 inches. You could eyeball it. And talk about all the reasons you think it's one or the other based on where the toothpick measures up to your ruler. But would you really be willing to bet $1000 just based on this wide confidence interval ruler? It's the same with election polls. If you have a +/- 3 point margin of error, that's a 6 point confidence interval, and it can't accurately measure, with statistical significance, a 5 point difference between candidates. So pollsters should just survey a few thousand more likely voters to put more marks on their election "ruler" so they can measure narrow differences with statistical significance. It really is an easy fix. And it is just annoying that polls continue to use such wide margins of errors in their polling. They know their polls are unlikely to match up come election day. And swings in numbers are very likely from poll to poll just due to random sampling error. But maybe that's exactly what they want, a swing to a number that supports their preferred news narrative.
Second, I'll talk about a new change in the poll aggregation averages at realclearpolitics.com. The new aggregate polls for Pennsylvania flipped from a slight Harris lead to a slight Trump lead. And in the no toss up map (where electoral votes are awarded no matter how small the advantage in the poll averages) Trump now leads with enough electoral votes to secure the presidency 281-257.
But aggregate poll averages introduce large amounts of error in the way they are calculated. So it pays to still be skeptical.
September 27th, 2024
A tale of two polls.
Realclearpolitics has two new polls for the battleground states of Arizona and Pennsylvania. In the USA Today/Suffolk poll of likely voters, Trump improved on the +2 Arizona poll average at +6 percentage points over Harris. The Bloomberg likely voter poll has Harris besting the +.9 Pennsylvania poll average to +5 over Trump.
With such large differences between the candidates in these two polls, especially considering how close other polls in the group were, it means that statistically, each candidate has moved into a lead in these two polls, right?
Let's look a little closer.
The Bloomberg poll surveyed 924 likely voters with a +/- 3.0 margin of error. And the USA Today/Suffolk poll surveyed just 500 likely voters for a huge +/- 4.4 margin of error. So in the Pennsylvania poll, Trump's support could be as high as 49%. And Harris' support could be as low as 48%. And in Arizona, Trump's support could be as low as 43.6%. And Harris' support could be as high as 46.4%.
So no, even with a 5 percentage point and 6 percentage point difference, neither poll result is statistically valid.
Why do they even run polls with such wide margins of error and inaccuracy? It's even worse for the Muhlenberg College Pennsylvania poll with a +/- 6 point margin of error. Meaning the confidence interval is 12 points wide. A candidate would have to be leading by more than 12 points before the poll could say with statistical significance that one candidate was ahead.
Surveying a few thousand more likely voters in each poll to improve the confidence interval would of course cost time and money. But with margins this wide, it's a wonder why the pollsters above even bothered to conduct a poll.
Such is the state of political polling at the current time.
If you would like to know more about margin of sampling error, you can read up on it.
September 26th, 2024
For today, I would like to draw your attention to an article NBC posted yesterday, "Chuck Todd: Are we at peak Kamala Harris?" This is a piece which is favorable to Harris so I know that there will be strong feelings on both sides about it. However, for the purposes of this site, it's important to note how Chuck Todd views the poll as evidence for his conclusions. And also learn something about poll margin of error in the process.
Chuck Todd states, "...the poll that I have the most confidence in: the NBC News survey."
So a few quick points:
The poll is over a week old in when it finished it's survey. So not being a recent survey is strike one.
The poll is a registered voter poll. If you are reading this site, you know that likely voters are what counts in election poll accuracy. Shame on Chuck Todd for putting his trust in a registered voter poll with all the available likely voter polls this close to the election. But he does work for NBC. So, not surprising. Strike two.
Chuck Todd uses sub-samples within the poll with very large margins of error to make claims about relative candidate support in these subgroups as if he can actually harvest specific support for a candidate with an accurate percentage. He writes, "Then there’s the issue of abortion rights. While it was an asset to Biden in polling against Trump, it appears to be a much bigger advantage for Harris. For instance, the NBC News poll showed her leading by more than 40 percentage points among white women with college degrees, a group of voters Biden won in 2020 but by a significantly smaller margin. Now, do realize, subgroups like white women with college degrees have margins of error that are at or close to double digits in the NBC poll. So even if there’s a 10-point error rate in this specific group, it would still mean Harris’ lead among this specific subgroup of female voters is 20 points..."
Statistically, he could have said, "Harris’ lead among this specific subgroup of female voters is 60 points..." and been just as wrong in his analysis. The margin of error gives you a confidence interval. Not a specifc support percentage. Inside of this interval, any differences between candidates are not statistically valid. And while the specifc 40 point difference quoted is statistically valid, Chuck todd can't then say that the real number is 20 points because of the margin of error. That's not what a margin of error is or how it is used. And this inaccurate subgroup comparison would be even worse if the polls he is using for comparison have equally large margins of error and confidence intervals for their subgroups.
It is much more accurate to say that the large confidence intervals of the poll mean that it isn't statistically valid for many statistics Chuck Todd is introducing, including the headline for the poll that Harris is leading Trump by 49%-44%. Which is outside the over 6 point confidence interval. Strike three. This is an example of the Chief Political Analyst for NBC needing a poll to support his narrative. But the poll he uses can't statistically support all his points. It would have been much better if the poll surveyed 10,000 likely voters instead of 1000 registered voters. But that's what it means when news people put narrative over actual poll evidence. They like the ambiguity and inaccuracy. They don't think they need a 1 point margin of error even in a close presidential race. They just go on with their narrative as if it's ok to draw conclusions with differences withing the margin of error.
Please jump here for an explanation of margin of sampling error.
September 25th, 2024
The American Spectator published an article titled, "Why Is Kamala Still Losing?". I know there are strong feelings about the topic of the article depending on if you support Harris or Trump. However, I'd suggest that the premise of the article is important if you want to understand the current state of election polling.
If you go to realclearpolitics.com, a website that averages polling data from various firms, you will notice boxes on the right hand side which break down and compare presidential polling data from 2016, 2020, and 2024.