Why we have to stop predicting for 2024
For those tempted to look at these polls or the current political environment to find out about the next general presidential election, all I can say is stop this. Neither the current presidential election nor the current political environment tell us much about what will happen in the 2024 general election.
Let’s start with the horse racing survey (i.e. the matchup between two candidates). There were seven previous election cycles prior to this one in which polls were conducted between an incumbent and his eventual challenger at this point in the cycle.
If you match the query and the final margin, the relationship is statistically insignificant.
The average difference between the electoral lead and the actual electoral lead is 8 points. In 2017, Biden had a lead of more than 11 points at this point. In the end, he beat Trump by nearly 4.5 points in the referendum – a difference of nearly 7 points from the early polls.
Note, however, that the average difference between survey and result in a small historical sample can greatly underestimate the potential differences in the future. Taking into account past differences and sample size (seven choices), the 95% confidence interval for how well surveys predict the outcome is +/- 30 points.
In fact, Harry Truman had a 28-point lead over Thomas Dewey in the late 1945 polls. He would win in 1948, but only with 4.5 points.
Either way you look at it, the horse racing poll doesn’t tell us much about how things will play out in 2024.
A big reason the polls don’t predict the outcome at this point is that we really have no idea what the political environment will be like on election day 2024.
The best way to measure the political environment, especially if Biden ends up becoming his party’s candidate, is by the president’s approval rating.
But if you look at a president’s net approval rating at this point and compare it with the result of where it ends up in the next presidential election, there is simply no connection.
For the 10 presidents since World War II seeking additional terms, there was an average difference of 29 points between their net approval rating at that point and at the time of the election.
Given that two of the last four presidents (Clinton and Trump) had net gains of about 10 points or more between that point and their re-election, it would be silly to bet that Biden will do the same.
You’ll also find that Clinton and Trump saw their parties swear in the midterms.
There are many presidents (such as Jimmy Carter in 1980 and George HW Bush in 1992) who lost their re-election even though their parties outperformed the historical average in the previous midterm elections. There are many presidents (like Clinton in 1996 and Barack Obama in 2012) who have been re-elected even though their parties have underperformed the historical average in the mid-term elections.
On the other hand, there are presidents whose party suffered minimal losses in the midterms (e.g. Richard Nixon 1970) and who were easily re-elected.
Conclusion: The current polls tell us about the current political environment before half-time. But the data we examine at this point in time, and even a year from now, doesn’t tell us much about what will come after that.