Replies

Thank you for your illuminating post. As someone who works with statistics, but is far from an expert, I can tell you that explaining them is almost as hard as understanding them.

I believe that you and Kevkrom are correct as to the main reason the exit polls were bad. Bad sampling.

Thank you for your illuminating post.

You're welcome, and thanks for the compliment.

I believe that you and Kevkrom are correct as to the main reason the exit polls were bad. Bad sampling.

I wouldn't even go so far as to say the exit polls were bad... after all, the adjusted data turned out to be quite accurate. The issue is that the raw data isn't very useful.

Example. Let's say you want to know who will win the 2004 election, so you poll 100 people. 55 prefer Kerry, while 45 prefer Bush. Ah hah, you think. Kerry will win.

Except looking at your sample, you notice that you happen to have polled 60 women and 40 men. You know that the electorate is divided 50-50 by gender. Looking closer, you see that of the 40 men, 35 preferred Bush while 5 preferred Kerry, and of the 60 women, 50 preferred Kerry while 10 preferred Bush.

Using this data, you "normalize" your sample. 87.5% of men prefer Bush, while 16.7% of women prefer Bush. Since the electorate is evenly divided by gender, you can average these figures and come up with a net support level for Bush of 52%, while Kerry gets 48%.

If you understand statistics, you will not be surprised when Bush wins the election. If, on the other hand, you're a DUmmie, you will howl conspiracy and point to the raw, non-normalized numbers as proof.