Reminder: betting markets have Bernie as the *most* electable candidate.

[andym]

63. So your numbers don't really mean that Bernie's chances are 58% of winning after the nomination

Last edited Sun Feb 23, 2020, 05:09 PM - Edit history (1)

Just that his chances of winning are greater than the current 34% -- that is correct.

Let me show you with a concrete example.
The problem of getting an accurate ultimate calculation is the differing sizes of pools of voters used in your calculations.

Let’s try some hypothetical numbers on for size (I think Bernie would actually do much better, but these illustrate the problem):

48 Democratic voters for Bernie ultimately (after no more Democrats running)
52 voters for Trump ultimately (after no more Republicans running)

[I've edited this OP to correct the misassumption that the percentage of votes is equivalent to the probabiity of success. Actually
the probability of success is equal to the probability of getting a majority (greaterthan 50%) given a random sampling of some size N. It is correctly determined by using a hypergeometric distribution (picking randomly without replacment) in which the one sums the number of all successes greater than 50%.
Here is a calculator https://stattrek.com/online-calculator/hypergeometric.aspx ]


Lets assume we are sampling half of the 48 Democratic voters (24) and then in the general election half the entire voting population of 100 (50) that I mentioned.

In this case using the calculator, Bernie would need only 26 voters out of the 48 Democrats (54%) to have a 61% chance of winning the nomination: population size 48, number of successes in population 26 (his total voters if everyone voted), sample size 24 (half of 48 Democratic the voters vote/are sampled for the nomination). Probability of getting 13 or more voters (13/24>50%) is 0.61= 61% (better than the Predictit 58%). So in the scenario Bernie needed to have 26/48 potential Democratic voters to have a 61% chance to win the nomination.

What about the general election? Let's again assume 50% of voters (this time all 100 voters vote)=50. Now we could choose the numbers so that Bernie wins the general election. But to show how one can't just assume winning the Democratic (or Republican) nomination proportionately increases the amounts in the general election, instead let's assume there are more voters who want Trump. Say 52 out of 100. Though Trump has a 93% chance according to Predict-it (the other 7% are assuming something happens to him presumably) these numbers would give him a 99% chance of winning his nomination.

So in this general election scenario, Bernie's chances of winning would be going down (not up): population size 100, number of successes in population = potential Bernie Voters 48, sample size of 50 (50% of everyone votes), number of successes in sample (majority of vote)= 26 (out of 50). The probability for Bernie to get 26 or more votes is 27% according to the calculator.

But according to your original post he could actually get 58 voters (58%), and that could eventually be true in reality. But there is no reason to believe that even if Bernie captured all of the votes of the other Democrats that his general electorate percentage would increase proportionately. It would likely increase, but limited by the number of people in the pool of potential Democratic voters.

The problem is the very different pools of voters over which your probabilities are calculated. That’s why you are receiving so much feedback.