It is possible that a sample size of 100,000 may be enough to draw a proper conclusion. It depends on the sampling criteria and methodology used. That is often why analysts will use a sampling error deviation (e.g. + or - 3%).
The problem with today's political polling methodology (and other polls) is that people deliberately skew the sampling to favor their views never reporting the bias of their own statistics. If a person polls 70% Democrats, then the polls will not be accurate (which is what happens today).
This is the original study - don’t know if it states how they made their sampling criteria and methodology. Perhaps is is something you might be interested in reading
https://phe.rockefeller.edu/news/wp-content/uploads/2018/05/Stoeckle-Thaler-Final-reduced.pdf