Pollsters were predicting a red wave mostly. Even left-wing pollsters like 538 were adjusting their initial projections. I am not an expert on polling nor am I a PhD in statistics, but I know enough to be dangerous.
In the aftermath, pollsters are going to try and figure out what went wrong.
In 2016, they all had Hillary beating Trump and when Trump won the Democrats accused Trump of stealing the election. Trump lost in 2020 and if you questioned the results you were a traitor.
The midterms certainly had some anomalies with them when it came to counting. Because it takes so long to count in states that are hotly contested, it invites criticism that everything isn’t on the level.
For example. I went to the Las Vegas Desert Breeze location to cast my ballot in the last hours of early voting. When I got there, a woman fell on the pavement in front, and inside a guy in a wheelchair was having a health problem. Fire alarms went off and everyone was pulled out onto the street including all election judges. We were outside for at least 15 minutes. No one was watching the kitty and I don’t know if there were remote cameras that were watching.
When my wife voted on the voting machine, it didn’t initially record her votes correctly. Good thing she checked. Going to paper ballots has to be done in the future because there are so many problems with machines.
Most probably everything was on the level, but because of the way America is with deep distrust between the political parties, there should be a speckle of doubt. By the way, if you are a Libertarian give it up. Infiltrate the Republican Party and take over like the Socialists did in the Democratic Party. You cannot govern unless you win elections.
One thing that occurred to me is that polls might no longer be random samples. Because pollsters have to call so many people to get a response, the probability chain changes. Doesn’t it become a different probability problem from the way pollsters initially set up the problem?
In 1948, pollsters missed elections because of the sample they got. They telephoned people to get answers and most people who owned telephones at the time were Republicans.
With extended early voting periods, mail-in ballots that have no control over them, and voter rolls that are not kept up to date, you can’t sample the group in the right way to replicate the potential event and make a prediction.
I think pollsters need to start using more powerful ways to analyze the electorate. Bayes theorem is one that all pollsters know about but I am not sure if they apply it to polls today. Or, should the problem be set up as joint probability? If you don’t know the difference, here it is defined simply.
Bayes Theorem
We know that,
P(A and B) = P(A)P(B|A) and P(B and A) = P(B)P(A|B)
When we equate this we will get, P(A)P(B|A) = P(B)P(A|B), then
P(A|B) = P(A) P(B|A) / P(B)
This is the Bayes theorem
It tells: how often A happens given that B happens, written P(A|B),
When we know: how often B happens given that A happens, written P(B|A)
and how likely A is on its own, written P(A)
and how likely B is on its own, written P(B)
In Machine Learning terms, Change A as Hypothesis and B as Evidence, then
P(A|B) = P(A) P(B|A) / P(B) becomes P(H|E) = P(H) P(E|H) / P(E)
This relates the probability of the hypothesis before getting the evidence P(H) — prior probability, to the probability of the hypothesis after getting the evidence P(H|E) — posterior probability. The factor that relates the two, P(E|H) / P(E), is called the likelihood ratio.
Bayes Theorem states that “The posterior probability equals the prior probability times the likelihood ratio”.
Joint Probability
Joint probability is the likelihood of more than one event occurring at the same time P(A and B). The probability of event A and event B occurring together. It is the probability of the intersection of two or more events written as p(A ∩ B).
Example: The probability that a card is a four and red =p(four and red) = 2/52=1/26. (There are two red fours in a deck of 52, the 4 of hearts and the 4 of diamonds).
Conditions for Joint Probability
One is that events X and Y must happen at the same time. Example: Throwing two dice simultaneously.
The other is that events X and Y must be independent of each other. That means the outcome of event X does not influence the outcome of event Y.
Example: Rolling two Dice.If the following conditions met, then P(A∩B) = P(A) * P(B).
What will happen if we find the joint probability of two dependent events?
Let Event X is the probability there are clouds in the sky and Event Y is the probability that it rains. Everyone knows that rain comes from clouds. So rain can only fall when there are clouds in the sky. That means the presence of clouds will influence the chances of rain, and that means these two events are NOT independent!
Joint probability cannot be used to determine how much the occurrence of one event influences the occurrence of another event. Therefore the joint probability of X and Y (two dependent events) will be P(Y).
The joint probability of two disjoint events will be 0 because both events cannot happen together.
One problem I had with the post-election analysis of the 2020 election was that assertions were made that because one county voted a particular way, another bordering county should have voted similarly. Often, that is not the case. We can cite lots of examples but if you are familiar with Chicago Lake County is not going to vote the same as Cook County.
The pollsters will not have this fixed by 2024 because I think it’s a hyper-thorny problem that requires quite a bit of money to fix. They also don’t have enough events that replicate a national election to model to experiment and get it right.
They can go through their past data and redo it using different methods to analyze and see if they get different results. But, I am not sure if that will make them better or not.
Probability is a funny thing. It is often hard to really wrap your brain around because you have to truly think differently. It’s not thinking logically like a math problem, but math is involved in the analysis!
Mollie Hemingway pointed out that pollsters are still geared towards election "day". We now have election season. Dems have mastered beautifully.
It's difficult to generalize about polling and polls because of the various sponsors and clients and the demands that they may make, or the level of confidence the client is willing to pay for. I think there is a substantial difference between what a well financed campaign receives in it's polling and what a news outlet gets from it's paid polling. Thanks to our public education system no one understands probability or standard deviation. Some of the polls may not be as wrong as they seem. If they predict a win by 2% with a standard deviation of 3% and it comes down to 500 votes the wrong way the poll wasn't really wrong. There is also the effect of the likely voter. In the recent past, polling firms that could narrow down to likely voters were often the closest approximation. In 2020 and especially in 2022 the likely voter has been replaced by the voter who's vote was harvested by a representative of the teacher's union, SEIU or some other activist group. This is a new challenge for polling firms, not to mention the opposition. We'll see if they catch up.