These intelligent questions about the (or,any poll's) poll's validity are reasonable and common.
I studied polling at the Kennedy School for a few years while at Harvard Business School. I have run political campaigns and employed top polling firms to provide tactical and strategic inputs.
The poll is fine. The margin of error is higher when samples are smaller. But the fact that a poll has a little higher margin of error does not sully the results. See the explanation below re margins of error.
Moreover, likely voter screens vary considerably in
severity---Registered to vote and strong interest in voting in the next election though never voted before?Voted in last election? Voted in last 2 elections? Voted in 2004 and 2008 but not in 2006? Any affirmative answer could pass a
particular likely voter screen. Typically, likely voters are more in touch and interested . They tend to expand in numbers closer to the election. They are generally the most informed repondent a pollster can hope for.
In this poll,I like the use of a likely voter screen---but understand that the
strictness of the screen could yield a sample of repondents much closer to merely
registered voters, than to wired people who vote in every single election.
Here is an explanation of margin of error:
What Is The Margin Of Error? Notice that our sample is an almost replica. There is a catch: we must make a trade-off when we choose to interview a thousand people instead of 200 million. That trade-off is the margin of error. Since we are talking to relativey few people, we can only say that our results are correct 95% of the time, give or take a few percentage points. For example, lets say that you see a CBS News story which says that 75% of the public has read a book in the past month. A graphic on the screen has a line that reads: Margin of Error +/- 3 percentage points. That means that if we asked all 200 million Americans the same question, we are 95% sure of getting a result anywhere between 72% and 78%. Since there would be just as much chance that the real figure was 75%, we report that figure. And our conclusion that Americans read a lot would stand regardless of that real figure. Sometimes you will see a story which says that a certain election or public opinion question is too close to call. That is due to the margin of error. Suppose Candidate A has 51% in our poll and Candidate B has 49% -- but the margin of error is three percentage points. Since the difference between the two candidates is smaller than the margin of error, each is as likely to win as the other. We cannot (and will not) say that Candidate A is in the lead. You will notice that when the margin of error prevents us from saying for certain who would win, we will not do so. Thats why we will always tell you the margin of error in stories about our polls on CBS News and in The New York Times so you can tell for yourself if our results are correct or if they are too close to call
Best regards--Blueprofessor