Rules for not bitching about polls

The UK Polling Report has a good rundown of the typical complaints about polls and the real reasons why those complaints are usually bogus.

I would also add, in regards to the margin of error and dead heats, that the media reporting of these MoE is usually faulty. Take the UKPR example of a 46%-45% result. The media would usually scream “dead heat” at this given it’s within the 3-pt margin of error. But what the media also screws up is that 46%-42% is also within the margin of error. We cannot be 95% confident that it’s not 43%-45% or even 49%-39%. We can be reasonably confident that the candidate at 46% is ahead, and can estimate that it’s a small lead, but we’re not extremely confident – there’s always a chance we’re wrong (and there’s even a built-in 5% chance that we’re dead wrong).

Only in a reported result is outside double the margin of error, say 47%-40% with a MoE of 3 pts, can we say we’re 95% sure the guy at 47% is ahead.

So, you see, polling is far from an exact science. That’s why campaigns rely on tracking polls to get updated numbers and trend lines. It’s also why some analysts aggregate or simply average out polling data to smooth out outliers and increase confidence in the results.

Just because video games and Hollywood give greater weight to individual polls in their political entertainment doesn’t mean any self-respecting political analyst should. But that doesn’t mean the results are completely bogus, either.