On the D side there are 370 delegates at stake between Texas (193), Ohio (141), Rhode Island (21), and Vermont (15). Ohio is currently the hardest outcome to predict. Signs are still pointing to a small Clinton win, but volatility in both the prediction markets (high) and the polls (less, but still present) makes the outlook murky. With at least 12 polls by 10 different pollsters in the past few days, most giving a 5-10 point lead to Hil, her lead seems safe. However, some things to consider:
- Repeat polls by SurveyUSA (solid track record) and Rasmussen (not so much) both show a 10-point Clinton lead falling to five points a few days later (both most recently taken on the 25th).
- No pollster has surveyed more than 862 “likely voters” and most are at n=600. At most, only about 0.01% of registered voters are being sampled. That would scare any rational statistician. Political pollsters get irrational quickly.
- The difference between the final outcome and the polls depends on whether any voting groups are being systematically underpolled.
For the most part, Obama’s tally margins over the polls have been greater than Clinton’s. In other words, if you look back at every state that has voted so far and compare final outcome with final polling averages, often the match was good (within a couple percent), but when it wasn’t, it was much more likely to underestimate Obama’s percentage and overestimate Clinton’s. This effect was particularly big in Georgia (Obama polled at 14 and won by 35), Wisconsin (Obama polled at 7 and won by 17), Tennessee (Clinton polled at 22 and won by 13), Alabama (polls tied and Obama won by 14), and to a lesser extent Oklahoma (Clinton polled with a 29-point lead and won by 24), even New York (Clinton polled at 20 and won by 17) and other states. Some states were right on (NJ, CT) and only one reversed the effect (in California Hillary polled by 6 and won by 9).
