About the polls...Modern polls are fundamentally flawed. They rely on online solicited paid survey takers resulting in biased, non-random samples. These polls lack the involvement of trained statisticians, instead being conducted by "political science" enthusiasts who often lack a deep understanding of mathematics and theoretical statist…
About the polls...Modern polls are fundamentally flawed. They rely on online solicited paid survey takers resulting in biased, non-random samples. These polls lack the involvement of trained statisticians, instead being conducted by "political science" enthusiasts who often lack a deep understanding of mathematics and theoretical statistics. Accurate results are crucial, yet these flawed methodologies cannot be corrected through weighting or data manipulation. In the end all you end up with is a non-random sample that represents paid survey takers not the US population. For instance take the Marquette Law School poll. To their credit Marquette Law publishes the actual sample sizes and response rates. Their February 2024 poll had a response rate of 2.07% with an ME of +/- 4.3. In real practice, done by a statistician, the actual margin of error is corrected for such a low response propensity by adjusting using the factor √(1/p-1) or ME *√(1/p-1) where p is the response rate. After adjusting for the abysmal response rate you attain +/- 4.3 * (6.88) or +/- 29.577. This is the accurate way of reporting results but don’t expect any of the pollsters to do this. Most if any even know how to do this. Remember how the polls predicted a win for Hillary Clinton? Well it just happened again. These inaccuracies highlight the unreliability of current polling practices.
Well, Ned, the truth is, political polling just isn’t what it used to be. Back in the 1970s, if a survey had less than an 80% response rate, we'd be up in arms about its quality. Nowadays? Phone surveys are lucky if they get even a 1% response rate. So, they’ve pivoted to the internet. But let’s be honest—most of them now rely on paid online survey takers. They call it using “panels,” and there are hundreds of these panels out there. I get invites to join paid surveys almost every week! It’s a mess and far from reliable. We haven’t cracked the code on accurate polling yet, but eventually, someone will.
About the polls...Modern polls are fundamentally flawed. They rely on online solicited paid survey takers resulting in biased, non-random samples. These polls lack the involvement of trained statisticians, instead being conducted by "political science" enthusiasts who often lack a deep understanding of mathematics and theoretical statistics. Accurate results are crucial, yet these flawed methodologies cannot be corrected through weighting or data manipulation. In the end all you end up with is a non-random sample that represents paid survey takers not the US population. For instance take the Marquette Law School poll. To their credit Marquette Law publishes the actual sample sizes and response rates. Their February 2024 poll had a response rate of 2.07% with an ME of +/- 4.3. In real practice, done by a statistician, the actual margin of error is corrected for such a low response propensity by adjusting using the factor √(1/p-1) or ME *√(1/p-1) where p is the response rate. After adjusting for the abysmal response rate you attain +/- 4.3 * (6.88) or +/- 29.577. This is the accurate way of reporting results but don’t expect any of the pollsters to do this. Most if any even know how to do this. Remember how the polls predicted a win for Hillary Clinton? Well it just happened again. These inaccuracies highlight the unreliability of current polling practices.
Exit polls used to be highly reliable and they made it hard to hide any cheating.
Well, Ned, the truth is, political polling just isn’t what it used to be. Back in the 1970s, if a survey had less than an 80% response rate, we'd be up in arms about its quality. Nowadays? Phone surveys are lucky if they get even a 1% response rate. So, they’ve pivoted to the internet. But let’s be honest—most of them now rely on paid online survey takers. They call it using “panels,” and there are hundreds of these panels out there. I get invites to join paid surveys almost every week! It’s a mess and far from reliable. We haven’t cracked the code on accurate polling yet, but eventually, someone will.