Yes. And then again ...

We seem to have some of the fundamentals confused downtown. Polling relies on random sampling, not random selection of campaign cliches:

Schuette leads Totten 38%-32%, according to the most recent Free Press poll. But with a 4 percentage point margin of error, the numbers could be much closer. And that gives hope to the Totten campaign, especially considering of the 600 people surveyed statewide 70% said they didn't know who he was.

Yes, the numbers could be much closer. On the other hand, there's an equally good chance that they could be much farther apart. The "margin of error" doesn't just apply to both proportions; it applies to both directions. And it isn't magic. It's a calculation. It expresses the range within which the sample value -- the proportion of respondents who say they're planning to vote for Candidate A or Candidate B -- accurately represents the population value, which is A's or B's support among all likely voters, or registered voters, or adults with landlines, or whatever it was you sampled.

In this case, there are about two chances in three that the sample is within 2 points of the population value* -- meaning that if the real split was 36-34, we'd have a pretty accurate poll. But it would be an equally accurate reflection of a 40-30 split.

There's another then-again to worry about. Voters who can't identify a candidate in a survey might come to support the candidate once they figure out who she or he is. And then again, they might not. If the story is meant to shed some light on what's often an obscure statewide race, it should stop spinning its wheels and get to work. 

* It's fractionally smaller in a race like this; the Totten vs. Schuette proportion is 32-38, but the Totten vs. not-Totten proportion is 32-68, so Totten's margin of error at 95% confidence is about 3.7 points. In a poll that reports lots of different races, it's easiest to calculate sampling error as if everything was split 50-50.