As I’ve written before, every time I review the crosstabs of the NYT/Siena College poll I find something dodgy. Sometimes it’s more than one thing—like over-sampling those with no college education, rural voters, and those in the south at the same time. Other times it’s simply over-sampling the south while generally being accurate in other areas. This time they are throwing all those other things overboard for a very simple over-sampling of conservatives.
According to the Pew Research Center 49% of registered voters are Democrats or lean to the Democratic Party, and a nearly identical share – 48% – are Republicans or lean to the Republican Party (from April, 2024). The “leaners” comprise 35% of the electorate. If we want to break it down by including an “independents” as opposed to a “leaners” category, according to Gallup the current break down (Oct. 2-12, 2024) is 31% Republican, 41% independents, and 28% Democrats. Of course this data is also based on polling and surveys so take it all with a grain of coarse sea salt.
Back to the dodgy sampling of the NYT/Siena poll: Way down in the crosstabs you’ll find “And just a few more questions for demographic purposes” and this is where I think they skew the poll. Respondents are asked to identify their political ideology and in the middle we have the moderates at 34%. If we combine Very Liberal and Somewhat Liberal we get 24%, and when Very Conservative and Somewhat Conservative are combined we get 38%! The moderates line up fairly well with both the “leaners” and the “independents” from Pew and Gallup, but the liberals are too low and the conservatives are too high. Higher up in the crosstabs is the question “If the 2024 presidential election were held today, who would you vote for if the candidates were” there is a 48% tie between Harris and Trump. With 38% of their respondents self-identifying as conservative this would mean a big number of “leaners,” or “moderates,” or “independents” are breaking for the MVP and only about 10% of that group is breaking for Trump. Their sample of only 24% liberals is only half the percentage who say they will vote for Harris. If this sample actually included approximately 33% of all three groups think of how that would change the results.
Polls don’t freak me out and they don’t make me complacent, but with good reason I ignore the results of this poll. How do you think the over-sampling of conservatives impacted the final results? Why do you think this poll often skews their samples to favor right-wing voters?
