I'm arguing it because it doesn't seem out of the question that it's a bad poll. If there is evidence that pollers are throwing out polls like this left and right, I guess that's something else.
BTW, just for giggles, I calculated Cruz's mean on this poll including all polls during this month. His score on this poll is 2.45 standard deviations above the mean for polls since 2/1. Assuming a normal distribution (although granted, I don't know enough about voting science to say you can), that would be a .7% chance of getting this result (or greater).
But remember there's a LV and RV poll, too.
And even a 1% number isn't proof against. Look, if I run 10,000 polls, I'm going to get some super funky polls in there, right? If I ran 10k polls and got this WSJ result, you wouldn't even think to question it. You would say it's a normal outlier. But because we only have like 15 polls in the last month, it looks super out of place, which is why you assume it's bad. Considering the reputation of the pollster, it's likely it's just an outlier. Outliers don't mean just on the edge of 2 standard deviations. They can be more than that and often will be.
To declare something not an outlier, I think you have to demonstrate within the crosstabs that something funky is going on. Too many old people assumed or too many white people or not enough catholics. Whatever it is. Like in 2012 when I said Gallup was oversampling the South which was skewing their results GOP.
The number itself isn't enough of an argument. We don't know how much herding other pollsters do. Or how many perceived outliers they never release.
Obviously, the WSJ poll looks to be at least an outlier. But to argue it's bad methodology can't be made based on the numbers alone. If I ran 10k perfectly fine polls, 1 of them will get Trump at 20 or so. It's the nature of the beast.
edit: FWIW, the PPP poll looks waaaay more suspect than this one.
And I really don't see the point of arguing this. If you want to argue it's a stupid poll, make the case in the cross-tabs and fine. If not, then it's just likely to be an outlier and move on. Who cares? Outliers are are not uncommon.
