I’m of course happy to see in the latest ABC News/Washington Post poll that President Bush seems to have reversed John Kerry’s momentum during a period in which Kerry should have been padding his lead: “High stakes convention.” I’m not particularly keen on this kind of horse-race poll so far in advance of the election. However, I haven’t seen anyone point out this poll’s methodology:
This ABC News/Washington Post poll was conducted by telephone July 22-25 among a random national sample of 1,202 adults, including 909 registered voters. The results have a three-point error margin.
If roughly 85 percent of registered voters vote (per Census Bureau analysis of November 2000 data), and this poll’s sample includes 25 percent unregistered adults, it might reflect the opinion of roughly 65 percent of likely voters — although it would not be a valid sample of them. Isn’t such a poll worth unusually little? Just asking.