This news is a bit old, maybe a week now, but American historian Allan Lichtman, who has correctly predicted every US presidential election outcome since 1984 using his "The Keys to the White House" model, has stated Joe Biden will win the election.

giggidy

His model is shit. It only predicts the popular vote not the electoral college.

His model also said that Trump is “not charismatic”. He is a leftist fuckup

He predicted Trump's win in 2016, so no, his model does not only predict popular vote.

Then how did he fail to predict 2000? Keep backpedaling ill wait

2000 was 20 years ago, you can improve a lot in 20 years.

Not true. He predicted in 2000 that Al Gore would win. When Gore lost, he pivoted to "my model is only predicting the popular vote". In 2016, however, it predicted Trump to win the popular vote, which he didnt. So no matter how you spin it, his model's predictions have failed in either 2000 or 2016.


On a more substantive note, his model heavily relies on the economy, but it was tuned based on data from normal economic states. But we are currently in a highly abnormal economic situation, one where many standard models no longer apply, and one that hadnt occurred in at least 90 years. Simply put, there is no reason to expect his model, which was tuned based on normal years, to still work in the current, highly abnormal situation.

This post was edited by Black XistenZ on Aug 11 2020 08:38pm

What do you mean how did he fail? Nobody claimed the model is perfect, and that race was so close that the supreme court basically had to appoint the president. If any model is going to get one wrong, that's gonna be it no matter what you're trying to predict.



2000 was so insanely close that I'm hesitant to call that incorrect in any meaningful sense. The recount data we got years later all but confirmed that Gore would have won if the supreme court didn't step in.

This post was edited by Thor123422 on Aug 11 2020 08:43pm

thats still a crazy reliable model, even if you exclude races like Obama-Romney where the winner was clear you still have someone doing insanely well on guessing coin flips.

Depends. When they write that his model "predicted" the winner in every race (but one...) of the last 120 years, do they mean correct classification on the full data set, or some sort of leave-one-out crossvalidation, or do they mean out-of-sample predictions for the respective next election period based on a model fit with all the data that was available until that point?

A perfect in-sample fit would be trivial with 13 'keys'/regressors/nodes, and LOO-crossvalidation is unrealistic since it would use information from future elections to "predict backwards in time". And then there's a ton of stationarity assumptions implied by this model which are definitely violated in the extraordinary 2020 environment. For example, the assumption that the importance of the economy for voters is the same in 2020 as in 2016... or in 1928.

It's crap that only looks good on the surface. Polling-based election models make far more sense, at least as long as they are set up by someone competent. All those fundamentals are typically captured by recent polling anyway, but polling can also catch structural breaks in how issues and fundamentals affect voting.

For example, in normal times, it is reasonable to assume that voters will blame the party in power if the economy is bad. A fundamentals model like Lichtman's has to work based on the assumption that voters will blame the incumbent for the economy no matter what, because that's how it was in all the years/data points their model was built on. But things might be different when an external shock like a global pandemic has caused an otherwise healthy economy to crater. Polling indicates that this is indeed true, the economy is still the issue where Trump polls best, and people still seem to trust him at least as much as Biden on the economy. So voters are seemingly not blaming Trump for the abysmal economy this year, which goes against all conventional wisdom and is something a fundamentals-only model cannot possibly account for.

This post was edited by Black XistenZ on Aug 11 2020 09:33pm

it's official. you cannot read and have a learning disability.

This post was edited by KungPeriodFu on Aug 12 2020 04:07am