Them don't understand polls
Them is direct object pronoun
“Them” is an object pronoun.
I like polls, them are good data.
Carl, are you okay?
Sometimes it's hard to make my point in just a few words, without getting too technical.
The first two sentences of this post (and the title) make it perfectly.
In my ample free time, I teach English to adults whose native language is not English. (I don't like saying English as a Second Language because many of them already speak 3 or more).
One of my challenges in teaching adults is figuring out how they learn. Some adults, like myself, like rules and charts. Even if the logic is inconsistent (see: English) I need to understand the rule.
To others, learning what an “object pronoun” is (or what direct and indirect objects are) just makes an achievable goal feel impossible.
You, English speaker, already know the differences - probably without realizing it.
You'd never say “I like polls, them are good data” or “I don't like polls, them are useless.”
But many non-native speakers would. Their grasp on the language isn't always as strong, and that's fine. Mistakes are a big, big part of language acquisition.
But now imagine people who spoke and wrote this way were the heads of English Departments worldwide, and insisted them they were right.
This is, without exaggerating, where the field of poll data/inferential statistics currently stands.
Here, in his recent book, G Elliott Morris (the one who just took over for Silver at FiveThirtyEight) outright says that a poll “predicted” something.
Nope. Them don't predict.
Somehow worse, even academic journals publish such nonsense:
Source: https://www.tandfonline.com/doi/full/10.1080/01621459.2018.1448823
No them didn't.
There are literally thousands, if not millions of examples of alleged experts who should know better being plainly, objectively, indisputably wrong about polls-as-predictions.
Pressed on the topic, those same experts (and other experts in the field, trying to defend them) will point to OTHER cases of them correctly talking about “polls are snapshots.”
Or, even - absolutely perfectly, as in this recent article:
Beautiful. Prestine. And accurate.
So obviously this author understands this fact, not simply regurgitating it?
Well, them is the direct object pronoun.
His article cites “accuracy” metrics from the same sources as above - based on, you guessed it, how well polls predict elections.
I will not assume students who are capable of regurgitating “them is the direct object pronoun” understand it when them misuse it.
Nor will I assume experts who are capable of regurgitating “polls are not predictions” understand it when them misuse data and analyze them as if them were predictions.
The fact that they can regurgitate what is true, as I did in the first line of this post, does not excuse misusing it, as I did in the second, and many times throughout.
In fact, the belief that you understand something, when it is easily demonstrated you don't, is far worse than admitting you don't understand it!
It's much easier to teach someone who thinks they need to learn than someone who doesn't. Trust me.
And that's the fight I'm currently fighting.
Anyone who has ever said any variation of:
"The polls predicted..."
"The pollsters declared as favorite..."
"The polls projected as winner..."
Does not understand how polls work.
And the above statements encompass 100% of the experts in this field, both those publishing and researching. I challenge anyone to point me to any expert who has never said any of those things.
And them would not say that if them understood what polls did.
If reading my sentences with the wrong pronouns give you a headache, consider:
“Them” in the above context is the wrong pronoun. An error that is grammatically incorrect but is still intelligible.
“Predicted” in the context of what polls tell us IS THE WRONG LANGUAGE.
To someone who understands what polls measure, using the word “predicted” is pure nonsense.
“I stood on the scale last week. It predicted what I would weigh today.”
You probably understand how nonsensical that statement is. Polls are no different.
Political polls have CONTAMINATED the minds of otherwise very smart people, and as a consequence the public has no chance.
I'm not just complaining, I have a fix
You may have guessed it from my English-teaching hobby, but I'm something of a linguaphile. Having words to convey meaning is…the whole reason we have language.
In statistics, a tool's accuracy is measured in comparison to the “true value.”
It happens, “true value” is taken to be understood, even when - as with polls - it isn't.
For a scale, the “true value” is not some future result. It's your weight the moment the scale takes it.
With poll data, it is incorrectly asserted that the ELECTION RESULT (not the population at the time the poll is taken) can be taken as the “true value”
Wrong.
When we don't have a term for something, we need to create one.
I created the term: Simultaneous Census.
Polls, as a tool, aim to ESTIMATE or APPROXIMATE, a Simultaneous Census.
What this means, is exactly as it sounds:
The poll estimates, or approximates (given the poll's margin of error) if you ask:
The exact same population
The exact same question
At the exact same time
With the exact same options
The result would be within the margin of error about as frequently as the confidence level suggests.
And that's it. Nothing to do with eventual result.
A good poll gives us good data to build a forecast, but the quality of the poll and the quality of a forecast are INDEPENDENT.
If you think the “Simultaneous Census” standard is too strict or theoretical, first of all, it doesn't matter - what matters is that it's correct.
But second of all, it's neither too strict nor too theoretical - the literal basis of the field's math, the margin of error, relies on all of these to be true.
Because our minds are contaminated by political polls, we cannot possibly comprehend how you could ask an entire population the same question at the same time.
But it's easy.
You just need a population that can't change over time, and whose future preferences you're not concerned with, but only their current ones.
Jacob Bernoulli, the discoverer of the “margin of error” used pebbles in an urn 300+ years ago. He sampled white and black pebbles from the urn - the whiteness and blackness of each pebble, since their color cannot change, and the contents of the urn are the population - are a Simultaneous Census.
That is, the measured “polls” from the urn can be compared to a Simultaneous Census.
I, much less importantly though more applicable to political polls, did so using mints in a box two years ago.
In most elections, the Simultaneous Census for any poll cannot be known. But that - not the eventual result - is what each poll tries to measure.
Understanding the differences between poll data in these two applications - which experts don't yet - is the first step to understanding what them tell us.