Politics: Poll Rant (part 1)

I promised it earlier and now seems as good of a time as any. As anyone who reads about politics online or in newspapers knows, the polls have been sort of all over the place lately. Some say Bush is up by 1 (essentially tied with Kerry); others say Bush is up by 8 (a complete Bush blowout). What do these polls mean? In short, nothing. Polling this election season is perhaps the most useless thing ever. Here's the start of a (hopefully short) list as to why:

1) National polls are meaningless due to the electoral college. We'll call this the "remedial civics" reason which should be obvious to anyone that remembers election 2000. Just how skewed can a national poll be? In the absolute worst case scenario, one can win the presidency by winning only 11 states: California, Texas, Florida, New York, Illinois, Pennsylvania, Ohio, Michigan, New Jersey, Georgia and North Carolina (for 271 total EV). Now this is unlikely happen thanks to some being staunchly Republican (i.e. Texas) and some being die-hard Democratic (i.e. California). In theory, however, someone can win the election by winning each of those states by 1 single vote. If that were the case, the popular vote total from those states would be around 29,933,487 for one candidate and 29,933,476 for the other. Now the same winning candidate could receive 0 votes in every other state, giving the losing candidate an additional 49,646,876 for a grand total of 79,580,352 votes. What would the national numbers show on election eve? The losing candidate would be ahead in the polls by 73% to 27%. All numbers used here were taken from projected turnout for the 2004 election from Race2004.

Moral of the story? The national numbers mean squat. You have to look at numbers state-by-state and electoral vote by electoral vote. That's why I recommend a careful look electoral-vote, Race2004, or Election Projection before deciding how your candidate is truly doing.

Next time on poll rant: Likely voters - Increased Poll Accuracy or Useless Superstition?