Allan Lichtman’s “13 Keys” election prediction system has a disputed track record and serious flaws. But it might still be worth learning about. A politics update by Jacob Cohen (@Conflux).
I'm pretty sure Allan Lichtman wrote long before 2000 that his keys predicted the popular vote (in his formal write-ups at least, he may not have made it clear in the more clickbaity articles), so I'm willing to give him 2000 as a correct prediction and put his track record at 9/10. It's not that surprising that he has such a good track record because:
1) As mentioned in my comment in the footnote, they keys all plausibly correlate with the actual results, so we should expect them to be right more often than not. I think the probability that he gets the average election right is higher than the 60% chance our markets give of him being right in 2028 - that probability is brought lower by the fact that 2028 will likely be another close election and therefore harder to predict than average.
2) There's selection bias in which pundits become popular. If he had a bad track record, we never would have heard about him, and instead we would have heard of Nalla Tilchman whose 25 Snap Guns to the Oval Office model predicted the correct results of every election except 2020 (She now claims that it was actually only supposed to predict the winner of Ohio). If enough people try to predict elections, we're going to end up hearing about the ones who do well purely by chance.
What really annoys me though, is that Lichtman's main claim to fame is, "He's the guy who predicted Trump would win in 2016," when that was the one year that his model was wrong! And his model was much more wrong than, say, Nate Silver's prediction that Clinton would win the Electoral College - she was closer to winning the EC than Trump was to winning the popular vote, and Silver made his prediction with very low certainty, while Lichtman seems to have very high confidence in his model's predictions.
The other thing that annoys me is that major news organizations report on his model so uncritically, completely ignoring the rather egregious fact that he completely changed what his model predicts after 2016 with no plausible theoretical basis, and the fact that he lied about this (He has falsely claimed that he actually changed it before 2016). Even if we take the 13 Keys completely seriously, they only predict that Harris will win the popular vote, which isn't a very bold prediction - everyone expects that already.
I totally agree! I meant to mention selection bias of pundits, but I might’ve forgotten to add it. And we do seem to be in an era of unusually close elections.
> He also says that the election was stolen in Florida, so Gore should have rightfully won the Electoral College as well.
This is quite probably true (in the sense that Gore should have won, not that there was a conspiracy or outright fraud), but also, it doesn't really matter whether it was true or not. Either way, the election was so absurdly close that it was clearly impossible to predict by any method.
I'm pretty sure Allan Lichtman wrote long before 2000 that his keys predicted the popular vote (in his formal write-ups at least, he may not have made it clear in the more clickbaity articles), so I'm willing to give him 2000 as a correct prediction and put his track record at 9/10. It's not that surprising that he has such a good track record because:
1) As mentioned in my comment in the footnote, they keys all plausibly correlate with the actual results, so we should expect them to be right more often than not. I think the probability that he gets the average election right is higher than the 60% chance our markets give of him being right in 2028 - that probability is brought lower by the fact that 2028 will likely be another close election and therefore harder to predict than average.
2) There's selection bias in which pundits become popular. If he had a bad track record, we never would have heard about him, and instead we would have heard of Nalla Tilchman whose 25 Snap Guns to the Oval Office model predicted the correct results of every election except 2020 (She now claims that it was actually only supposed to predict the winner of Ohio). If enough people try to predict elections, we're going to end up hearing about the ones who do well purely by chance.
What really annoys me though, is that Lichtman's main claim to fame is, "He's the guy who predicted Trump would win in 2016," when that was the one year that his model was wrong! And his model was much more wrong than, say, Nate Silver's prediction that Clinton would win the Electoral College - she was closer to winning the EC than Trump was to winning the popular vote, and Silver made his prediction with very low certainty, while Lichtman seems to have very high confidence in his model's predictions.
The other thing that annoys me is that major news organizations report on his model so uncritically, completely ignoring the rather egregious fact that he completely changed what his model predicts after 2016 with no plausible theoretical basis, and the fact that he lied about this (He has falsely claimed that he actually changed it before 2016). Even if we take the 13 Keys completely seriously, they only predict that Harris will win the popular vote, which isn't a very bold prediction - everyone expects that already.
I totally agree! I meant to mention selection bias of pundits, but I might’ve forgotten to add it. And we do seem to be in an era of unusually close elections.
> He also says that the election was stolen in Florida, so Gore should have rightfully won the Electoral College as well.
This is quite probably true (in the sense that Gore should have won, not that there was a conspiracy or outright fraud), but also, it doesn't really matter whether it was true or not. Either way, the election was so absurdly close that it was clearly impossible to predict by any method.