Not a statistical tie

Politico reports:

A new poll of Georgia voters finds Donald Trump and Hillary Clinton nearly tied in a general election matchup.

The Atlanta Journal-Constitution poll shows Trump with a 4-point lead over Clinton, 45 percent to 41 percent, which is within the poll’s 4.26 percentage point margin of error.

The sentiments expressed by independents further contribute to the statistical tie between the two presumptive nominees.

It isn’t a statistical tie. I do wish media wouldn’t use that term. It is good media point out a lead may be within a margin of error, but not good when they suggest that it is effectively a tie.

For a poll of 822 voters, the chance that Trump is actually leading is 89.2% and the chance Clinton is leading is 10.8%. This is less than the normal 95% confidence interval so Trump is not necessarily leading – but 89% probability is not a tie.

50% is a tie and 95% is statistically significant. I’m not sure what the term is for probabilities between 50% and 95% but it is not a tie – ie a 50.1% chance you are leading is not the same as a 94.95 chance you are leading.

Comments (18)

Login to comment or vote