Friday, October 19, 2012

A Technical Note on Polls – Why the Conflicting Results

It’s Statistics

Trying to determine the state of the Presidential race from national polling is confusing these days.  There are a large number of national polls, many of whom use respectable methods.  But polling is a sampling process, and polls are err to what is called sampling error.  That refers to the fact that the sample may not, and in fact will not be completely representative of the population as a whole.

Sampling error does not matter when a poll is not close.  If the Presidential race were truly say 55% for Mr. Obama and 45% for Mr. Romney, or 44% for Mr. Obama and 56% for Mr. Romney, a sampling error of a few percentage points would not matter.  The polls will still show the candidate that is leading with a big lead, and the candidate that is behind with a large deficit.

But when the race is close sampling error plays a big role in polls.  If the actual race for President at this time is, say 49.2% for Mr. Obama and 48.9% for Mr. Romney, then sampling error will play havoc with the polling.  An error of just two percentage points will distort the polling results and show one candidate or the other with a large lead that in reality does not exist.

This is why Nate Silver, the acknowledged expert in this area is finding contradictory polling results and why his mathematical modeling of the race showing Mr. Obama a clear favorite is in question.

But other national polls published on Tuesday were not in agreement with the Gallup and Public Policy Polling numbers. Rather, three of the six national polls published on Tuesday had Mr. Obama leading the race. The same three polls also had Mr. Obama improving his numbers from the previous edition of the same survey, while the other three had him declining.
This is a curious distribution of polls. While on average the polls showed almost an exact tie in the race, none of the individual polls did so, each lining up on the side of Mr. Obama or Mr. Romney.


In this type of race quantitative analysis is just not effective, the sampling error of the polling is too dominant and one cannot trust the results.  It is for this reason that qualitative analysis is required.  And that analysis says the Mr. Romney is winning, subject to any polling that reflects the second debate. Or the fact that Mr. Romney likes his women in binders.


No comments:

Post a Comment