Sources of Error

As Robert Waldman says it's almost certain that model error rather than sampling error is the main cause of inaccuracy in public opinion polling. You've especially got to wonder about that in something like the American Jewish Committee's poll of Jews (via Laura Rozen). The poll, showing Kerry leading 69-24 (better than he was doing in the last AJC poll, but worse than Gore did in 2000 according to the exit polls), is said to have "a margin of error of plus or minus three percentage points" based on its sample size of 1,000 Jews. That's nice, but the real problem here has got to be that there's some indeterminacy around the question of what we mean by "Jew" (take my friend Jeff Baum -- Jewish dad, shikse mom, Hebrew School, bar mitzvah, doesn't attend services or keep kosher) so we don't really know what this data is referring to, and it's not clear how you would obtain a representative sample of Jews since you can't use standard techniques like demographic weighting and so forth. Nor, of course, is the accuracy of the 2000 exit poll beyond dispute. And so forth. Nevertheless, the brutal reality is that an awful lot of American Jews seem to be in the grips of the anti-semitic conspiracy theories that, as we all know, are the only motive people have for opposing the president.