What I’ve come to learn however is that polls are problematic for another reason as well: Each pollster makes assumptions about the ultimate makeup of the electorate. Assumptions about who is going to vote and who isn’t. Assumptions about how many Democrats, how many Republicans and how many Independents are to be included in each of the polls.

Example: Out of one hundred people, let’s say the pollster decides upon the following demographic break-down: 38% Democrat; 32% Republican; 30% Independent. Each of these three groups have voting tendencies and are likely to vote for their corresponding candidate. Let’s say that the Democrats vote for Obama, Republicans for McCain, and Independents split 60:40 to McCain.

In the above example, we end up with a fifty-fifty tie. But were our assumptions correct? Does that part of the electorate which is actually going to vote break down at 38-32-20? Simply polling one more Democrat and one less Independent results in an Obama win.

Past polls have tended to be fairly inaccurate as it is. This year’s election makes things even more problematic. The Obama candidacy has found unprecedented support from blacks and the young, two groups who in the past have not voted at the same rate as other ethnic and age groups. How to account for these expanded voter rolls?